Not signed in (Sign In)

Not signed in

Want to take part in these discussions? Sign in if you have an account, or apply for one below

  • Sign in using OpenID

Site Tag Cloud

2-category 2-category-theory abelian-categories adjoint algebra algebraic algebraic-geometry algebraic-topology analysis analytic-geometry arithmetic arithmetic-geometry book bundles calculus categorical categories category category-theory chern-weil-theory cohesion cohesive-homotopy-type-theory cohomology colimits combinatorics complex complex-geometry computable-mathematics computer-science constructive cosmology deformation-theory descent diagrams differential differential-cohomology differential-equations differential-geometry digraphs duality elliptic-cohomology enriched fibration finite foundation foundations functional-analysis functor gauge-theory gebra geometric-quantization geometry graph graphs gravity grothendieck group group-theory harmonic-analysis higher higher-algebra higher-category-theory higher-differential-geometry higher-geometry higher-lie-theory higher-topos-theory homological homological-algebra homotopy homotopy-theory homotopy-type-theory index-theory integration integration-theory k-theory lie-theory limits linear linear-algebra locale localization logic mathematics measure-theory modal modal-logic model model-category-theory monad monads monoidal monoidal-category-theory morphism motives motivic-cohomology nlab noncommutative noncommutative-geometry number-theory of operads operator operator-algebra order-theory pages pasting philosophy physics pro-object probability probability-theory quantization quantum quantum-field quantum-field-theory quantum-mechanics quantum-physics quantum-theory question representation representation-theory riemannian-geometry scheme schemes set set-theory sheaf sheaves simplicial space spin-geometry stable-homotopy-theory stack string string-theory superalgebra supergeometry svg symplectic-geometry synthetic-differential-geometry terminology theory topology topos topos-theory tqft type type-theory universal variational-calculus

Vanilla 1.1.10 is a product of Lussumo. More Information: Documentation, Community Support.

Welcome to nForum
If you want to take part in these discussions either sign in now (if you have an account), apply for one now (if you don't).
    • CommentRowNumber101.
    • CommentAuthorAndrew Stacey
    • CommentTimeJan 27th 2011

    I’ve done the two diagrams as proper SVGs and sorted out the labels on the down arrows.

    I’ve copied it over to Leinster2010 now. That just leaves the wiki-links from my list.

    • CommentRowNumber102.
    • CommentAuthorTom Leinster
    • CommentTimeJan 27th 2011

    I'm awed by, and extremely grateful for, the work that's gone into this. But for the moment I'm going to hold back the full force of my thanks, and concentrate on details.

    It looks great to me in almost every way. Here are some exceptions:

    1. There's a problem with the diagram that accompanies the (wonderfully redrawn) section "cartoon", on this browser at least (Firefox 3.6.13 on Linux). Here's what I see. Unfortunately I don't know how to fix this myself.

    2. Citations. The citations used in the nLab page are my own private keys. Now, I didn't use any embarrassing nicknames, so I don't really mind. But it's not ideal: I'd never intended those keys to appear in the text, and I don't really want to inflict my own not-very-systematic naming on the world. So I'll see if I can figure out how to change them now.

    • CommentRowNumber103.
    • CommentAuthorAndrew Stacey
    • CommentTimeJan 27th 2011

    In 1, are you referring to the “could not include SVG …” bits? If so, that’s easily fixed: I just need to copy over the arrows from doriath to the main nLab. I thought I’d done all of the ones that your article used but clearly I missed these ones. If you’re referring to something else, please clarify.

    As for 2, go ahead!

    • CommentRowNumber104.
    • CommentAuthorTom Leinster
    • CommentTimeJan 27th 2011

    Some more comments:

    1. Citations. I re-did them, but something mysterious is going on with one particular item: Lawvere (1964). There's one reference to it (search for the first occurrence of "ETCS"), and clicking on it does nothing. If you look at the html of Leinster2010, you see that something's wrong with that item in the reference list. But I can't figure out how to fix it.

    2. Numbering. Is there a way of making all the different theorem-types (Definition, Examples, Theorem, etc) be numbered in sequence, as almost everyone does in Latex? E.g. if the first item of the paper is a definition, the second a lemma and the third an example, we want Definition 1, Lemma 2, Example 3, not Definition 1, Lemma 1, Example 1.

    (For my short paper I don't mind numbers of the form "1" rather than "1.1" or "1.1.1", but I suppose other authors might want that in the future.)

    Sorry if this has been discussed before.

    • CommentRowNumber105.
    • CommentAuthorTom Leinster
    • CommentTimeJan 27th 2011
    • (edited Jan 27th 2011)

    @Andrew (103): yes, that's what I was referring to. By the way, what is doriath? The name of a computer?

    • CommentRowNumber106.
    • CommentAuthorUrs
    • CommentTimeJan 27th 2011

    Andrew,

    concerning wiki-links and since you have already shown this amazing effort at automatizing the process: would it be very difficult to have a script simply check each string in the text for if there is an nnLab entry with that name or redirect, and insert double square brackets suitably, if so?

    • CommentRowNumber107.
    • CommentAuthorTom Leinster
    • CommentTimeJan 27th 2011

    Urs, I've been thinking a bit about this possibility (and I recall seeing a little discussion of it on one of these threads).

    The first thought is that wikilinking is really something we want to take advantage of...

    The second is that it if linking text has any visual impact -- i.e. if it's underlined or in a different colour -- then things could quickly get out of control. If every occurrence of "category" or "space" or "object" is highlighted then it'll be a visual disaster. Of course, if there's no visual impact then it doesn't matter (except perhaps for memory and speed of rendering). But I think Mike had reasons for wanting links to be visible.

    The third is that there may be the odd ambiguity that needs human intervention: e.g. consider the different meanings of "lattice" or "representation".

    • CommentRowNumber108.
    • CommentAuthorUrs
    • CommentTimeJan 27th 2011
    • (edited Jan 27th 2011)

    Right. This is actually a general issue that we also once talked about: it would be good to have the wiki-links be less visible in a way. A year ago or maybe longer we had discussions about this and played around with CSS, but nothing did really matierialize.

    I’d be in favor of this generally, for the whole nnLab: that we change the appearance of the links to make them have less visiual impact – while still being recognizable, of course.

    The first and foremost is, I think, that we should get rid of the underline. That’s too much. A little shade in color should be the right thing, maybe with something more when moused-over.

    This is easily possible by playing with the CSS. I once had this implemented on my personal web, but I have abondoned it for some reason. I forget which reason, though. :-)

    • CommentRowNumber109.
    • CommentAuthorAndrew Stacey
    • CommentTimeJan 27th 2011

    Tom, I’ll have a look at the citation issue and the numbering. I now understand how the numbering works so think I can fool it into behaving how I want it to behave.

    Doriath is one of the “webs” at the nLab. It’s like a Sandbox only bigger. It’s where the working copy of your article was for a while before I just moved it (note to self: I should put a banner on that copy redirecting to the nLab version). In short, it’s for doing experiments that are just a bit too big for the Sandbox. Take a look at HomePage (doriath) for more information. (And if you’re curious as to the name, I suggest you ask someone in your department who recently became a father if he recognises it. If he doesn’t, try telling him that there’s another of these webs here called “eregion”.)

    The CSS can be played with to our collective hearts’ content. There are extensions for firefox that allow you to play with it “live”, as it were, to see the effect instantaneously. The main issue for me is a lack of imagination and an unwillingness to spend ages doing little tweaks followed by huge arguments while everyone disagrees over the exact shade of blue used (on re-reading, that sounds a little harsher than I mean it to! I just mean that there are better uses of my time.).

    I can also write a script to put in all the wiki-links. I pondered that at the start of the conversion process. I’m hesitant about doing so, though. I think that this is really where we can truly “add value” and that putting in the wiki-links is probably the most important part of the whole process. Important enough that it ought to be done right, and that means not by computer but by a person. I don’t think that simply linking every possible word helps. I think, especially with an actual article like this, that there should be some thought as to where the links could actually be helpful and where there’s little chance of them being used. As a trivial example, it seems senseless to link every use of the word “topos”, but obviously a link at the start would be good.

    • CommentRowNumber110.
    • CommentAuthorUrs
    • CommentTimeJan 27th 2011

    The main issue for me is a lack of imagination and an unwillingness to spend ages doing little tweaks followed by huge arguments while everyone disagrees over the exact shade of blue used (on re-reading, that sounds a little harsher than I mean it to! I just mean that there are better uses of my time.).

    Certainly, don’t waste time on this. Just choose your favorite shade of blue or green and force it upon us all!

    it seems senseless to link every use of the word “topos”, but obviously a link at the start would be good.

    There is an argument to be made for really hyperlinking every single keyword. We just recently had a discussion about this on the nnCafé, where somebody suggested that wiki’s ought to be programmed to behave this way, and somebody boldly claimed that on the nnLab we are already trying to approximate that goal.

    Because there is this idea that on a wiki you have the advantage that you do not have to read the text starting at line one and proceeding linearly, so that you are sure you see the first and only hyperlinked occurence of any keyword. Instead, you can jump to any one statement in the midlle of the action that you need, and rest assured that for every keyword which you need help about that help is only one mouse click away.

    Of course, what you really want (I think) should be available, too: some links should be part of the author’s message. They can be highlighted. If the author really and explicitly wants to say: “see here for more information”, then there should be a special link (maybe just boldface would do, or the like) indicating this.

    • CommentRowNumber111.
    • CommentAuthorEric
    • CommentTimeJan 28th 2011

    Links on the nJournal could be there but not visible. I’d prefer not seeing the links in a formal article but feeling comfort knowing they are there.

    • CommentRowNumber112.
    • CommentAuthorMike Shulman
    • CommentTimeJan 28th 2011

    What about a “semi-automated” interactive link-adding script? I mean, a script which displays to the user the links it’s planning to add automatically, and allows him/her to confirm or deny each of them? As a command-line script, it could display each potential link with some context one by one and let the user bounce on the y/n keys (with maybe another option saying to make a link but to a different page than to the one guessed by the robot. The robot could even guess some other potential options and let the user pick one of those with a couple keystrokes, to make it quicker than typing out a whole different page name). Or as a graphical interface, it could display the text with all the potential links highlighted, and let the user click on ones that should be rejected or redirected. Either way would be quicker for the user than going through and adding them all by hand, but with less risk of adding wrong links than a fully automatic script.

    I am in favor of making wikilinks less obtrusive with CSS; I’m happy as long as they are indicated visibly in some way, but it can be as unobtrusive as we like.

    • CommentRowNumber113.
    • CommentAuthorTom Leinster
    • CommentTimeJan 28th 2011

    @Andrew (109): thanks.

    Re visuals: I sense a discussion that could go on for literally ever. All the same, here's my 2 cents.

    I said something like "if we link everything that we could, and linking has any visual impact, it'll be a visual disaster". When I say any visual impact, I really mean that literally. That is: in my opinion, if we link everything for which a wikilink exists, and we wish to avoid visual disaster, the links really must look exactly like ordinary text.

    To see what I mean, take a random paragraph from any category theory paper and observe how many words could be wikilinked. (The problem will get worse as the nLab accumulates more entries.) We could be talking about something like 25% of the text. It would make for a very bumpy read, even with the most subtle colour change.

    I can see two objections to links being visually undistinguished.

    First, you don't know what you can click on. Well, I think you'll learn very soon that the nLab is pretty comprehensive and most categorical terms are clickable. With mathematical terms that aren't categorical, it's harder to predict, but I think you just have to hover your mouse over and see.

    Second, sometime you (or I, at least) want to click on a neutral point on the page in order to focus the computer's attention on that window (e.g. so that you can use the cursor keys to scroll). That is a drawback, but I think only a minor one. Making that mistake won't take you inadvertently to some nasty website; you'll just accidentally see the definition of category or something. And you'll learn soon not to do that.

    I made a similar decision a while ago with hyperref. I think it's hyperref's default behaviour to put clickable links (e.g. the "5.2" of "Lemma 5.2") inside a red box. It's the default when things are processed through the arXiv, anyway. I think this is pretty disastrous: some pages are a sea of red boxes, and personally I find it very off-putting to read. So I now have my "5.2" looking exactly like ordinary text. Yes, you lose a little bit in terms of not informing the reader that they can click on it. But I think you gain much more in readability.

    • CommentRowNumber114.
    • CommentAuthorUrs
    • CommentTimeJan 28th 2011

    I am in favor of making wikilinks less obtrusive with CSS; I’m happy as long as they are indicated visibly in some way, but it can be as unobtrusive as we like.

    How about having wikilinks only indicated when moused-over?

    Maybe that is something worth trying.

    • CommentRowNumber115.
    • CommentAuthorEric
    • CommentTimeJan 28th 2011

    I agree with Tom. Links should be invisible generally.

    I also like Urs’ suggestion to make them visible when moused over.

    • CommentRowNumber116.
    • CommentAuthorAndrew Stacey
    • CommentTimeJan 28th 2011

    Maruku together with CSS should be powerful enough to do both visible and invisible linking. I’d have to experiment a little since Maruku doesn’t handle the wikilinks so we may have to be sneaky about it.

    I’ll have a think about the script.

    • CommentRowNumber117.
    • CommentAuthordomenico_fiorenza
    • CommentTimeJan 28th 2011
    • (edited Jan 28th 2011)

    I agree with this, too. My ideal would be that we link everything (to allow the “every point is a good point where to start reading” principle as remarked above) but with links that are invisilble unless moused over. This is a quite drastic change of habit for the reader, since there will not be links seductively saying “Hey, follow me ;)” but I find this would be a positive change. Most of the time, when reading a text with highlighted links I have to fight against the curiosity of following them. For instance if on an nLab page I find “Let 𝔤\mathfrak{g} be a Lie algebra..” I can’t avoid thinking “mmm.. yes, I know what a Lie algbra is… but let’s have a look at what we have on nLab on this, I could learn something new.. or I could spot something which needs fixing, or..”. so I’m distracted and in a sense have to decide to remain on the page I was originally intersted in. instead, if the link is hidden, only when I will feel that I do not know what “Lie algebra” is referring to I will try to see if there is a page for that.

    • CommentRowNumber118.
    • CommentAuthorAndrew Stacey
    • CommentTimeJan 28th 2011

    We should, however, have a big announcement at the top saying “Hover your mouse over a term to see if it is linked to an explanatory page.”

    • CommentRowNumber119.
    • CommentAuthorUrs
    • CommentTimeJan 28th 2011

    We should, however, have a big announcement at the top saying “Hover your mouse over a term to see if it is linked to an explanatory page.”

    We could add that just to the HomePage, which is also the page that explains how to openn the context boxes.

    • CommentRowNumber120.
    • CommentAuthorAndrew Stacey
    • CommentTimeJan 28th 2011

    Actually, I’d prefer a discrete link entitled something like “Useful information for readers unfamiliar with the nWhatever” on every page. But I agree with the basic point of putting it on a different page.

    • CommentRowNumber121.
    • CommentAuthorAndrew Stacey
    • CommentTimeJan 28th 2011

    Thinking about the script, what would the rules be? If it’s to be interactive then we’d want the rules to be pretty loose as it’s much easier to just hit ’n’ than to go back and add in links afterwards. But clearly there’s a tipping point where it becomes just plain irritating.

    As an example, the word ’category’ should clearly match something, but should it only match category or should it bring up a list of all pages with ’category’ in their name? Clearly, longer matches should be preferred over shorter matches, so (say) “category of semi-regularised normal abelian groupiods” comes higher than “category” in the list (assuming that’s what was matched).

    So I’m envisioning something that starts with a sentence (that being the longest possible thing to match) and works backwards, removing a word at a time and looking for a match. When it finds a match with a sufficiently high score, it presents all the matches with a certain score (which may be lower than the first one, after all if there’s a match at 10% and one at 9% then both should be presented). If none is acceptable, it continues with the process of losing a word at a time until finally either a match is accepted or the start of the sentence is reached. Then the first word is dropped and the process begins again. (To speed it up a bit, I suppose the process could short-circuit if there is no page name (or redirect) that starts with the first word).

    What test for matches should be used? Substring (with whitespace collapsed to a single space)? Substring ignoring punctuation? Substring ignoring mathematics? Balanced or not? (By which I mean that if the link text is a substring of the page name then that doesn’t necessarily score the same as the page name being a substring of the link text). Any other suggestions?

    • CommentRowNumber122.
    • CommentAuthorAndrew Stacey
    • CommentTimeJan 28th 2011

    Right, I’ve done the missing arrows, fixed the link to Lawvere’s anchor (an obscure thing about lines starting with a *), and done the counters. The problem with the counters is that now all the environments are typeset with italic body which is obviously wrong. That’s fixable, though, with a little CSS tweaking.

    • CommentRowNumber123.
    • CommentAuthorMike Shulman
    • CommentTimeJan 28th 2011

    Well, I’m still not convinced by completely invisible links, but it looks like I’m pretty unanimously outvoted, so I won’t make any more fuss. I’ll certainly get used to it; perhaps I’ll even come around to like it. (-: I’ll certainly be happier if they become visible when moused over (what a curious verb…). (Re: hyperref, I agree that red boxes are overkill, but I get slightly annoyed when I read a hyperref’ed paper with no visual indication at all, since there’s no way for me to tell that it is hyperref’ed, or where the links are, without experimenting by clicking on stuff.)

    • CommentRowNumber124.
    • CommentAuthorTobyBartels
    • CommentTimeJan 28th 2011
    • (edited Jan 28th 2011)

    I want to know about links; I don’t want to have to go hunting them with my mouse. They certainly can be less obtrusive that blue underlines, but I want to be able to see them and identify their individuality.

    For links in the main Lab, I would very much like: some visual indication of each link (no mouse-over or anything else required), including a visual element that spans the entire link (such as an underline), but the latter feature could wait for a mouse-over to appear.

    For links in the nProceedings web, however, things could be different. Especially if every occurrence of “object” or “topos” is linked automatically, perhaps I could get used to invisible links in this limited context. But I’m also pretty sure that I would find them damned annoying at first, if I came upon an online journal like this.

    Anyway, as long as you don’t mess with the nLab, you should probably all do what you like with the nProceedings.

    • CommentRowNumber125.
    • CommentAuthorTobyBartels
    • CommentTimeJan 28th 2011
    • (edited Jan 28th 2011)

    Thoughts on automated linking:

    I have nothing to improve on the basic algorithm in Andrew #121, if it’s not too slow.

    It will be even more important to have redirects of plurals and the like. The linking algorithm itself should probably be case-insensitive (since we rarely make such redirects and sometimes shouldn’t), unless it gets matches in different case, in which case it can rank them to match the case of the text.

    Adjectives will require a lot of user intervention; probably there’s no way around this without teaching the algorithm about the various fields of mathematics, which doesn’t seem feasible. We could also create a lot of adjective pages and have the robot link to them by default, but I don’t really like that; any such link ought to be corrected by a human.

    We may also want to come up with some exceptions, such as “the”. (Those could all link there, but do we really want that?)

  1. I say: let’s try it. I think it’s most a matter of habit, and that when one gets used to the idea that in the text one is reading everything is linked then to guide the reading it will not be “what is linked and what is not” but “what I’m interested in here and now and what I’m not”. So lets’ have a try on nPub, if it won’t work we can always change.

    • CommentRowNumber127.
    • CommentAuthorEric
    • CommentTimeJan 29th 2011

    Another idea might be to have a different “view” like the “print view” that allows you to easily toggle between invisible and visible links.

    • CommentRowNumber128.
    • CommentAuthorMike Shulman
    • CommentTimeJan 29th 2011
    • (edited Jan 29th 2011)

    If links are generally invisible, can there be some way for the author to designate some links as visible? Or, better, vice versa? I really don’t trust someone not familiar with wikis to find out the linkedness of everything, even with a comment somewhere that “you should try clicking on things.” People often don’t read the instructions, and the fact is that the current convention on the Internet is that links are visually identified.

    Edit: I really think this would be the ideal solution. I do agree that if we’re going to link everything, then it will look pretty untidy for reading if all links are visible. But I do think that some links, like the first time a concept is introduced, or when a definition from way earlier is referred back to, should be visible, to specifically encourage the reader to follow them.

    • CommentRowNumber129.
    • CommentAuthorEric
    • CommentTimeJan 30th 2011

    It would be awesome if you could modify the appearance of the link in the markdown syntax itself, e.g.

    [ [invisible::nlab:functor]]

    • CommentRowNumber130.
    • CommentAuthorAndrew Stacey
    • CommentTimeJul 22nd 2011

    Now that Tom’s edited the original in line with the referee’s report, I’ve started work on converting the final version to the instiki format. As is now standard, I have a completely new method! (Actually, it’s not completely new: in writing the Perl interpreter for TeX, I had to write a “conversion” package. I then realised that that would work with honest TeX, so now the system is to use TeX itself to do the conversion.)

    The latest version is at converting latex to itex (doriath). At the moment, that is almost exactly what is produced by the script - the only modifications are a few linebreaks (but as I don’t really understand linebreaks, I haven’t automated that aspect yet). There’s one outstanding error, where a definition environment doesn’t get properly converted, but it doesn’t affect the readability.

    I believe that the True Version of the paper is on the arXiv. Also, the previous version is at Leinster2010.

    At this point, I could do with some fresh eyes to point out errors and weird behaviour. So please take a look and tell me what looks wrong.

    (Note that I haven’t done anything about wikilinks yet.)

    • CommentRowNumber131.
    • CommentAuthorUrs
    • CommentTimeJul 22nd 2011

    Please let me know once we have a decent nnLab version of the published article. I should want to announce the event of the first refereed nnLab publication on the nnCafé.

    • CommentRowNumber132.
    • CommentAuthorTom Leinster
    • CommentTimeJul 26th 2011

    Thanks very much, Andrew. I'm just back from Category Theory 2011 in Vancouver, and will check through this as soon as I can. To confirm: yes, the true (i.e. accepted) version is on the arXiv.

    • CommentRowNumber133.
    • CommentAuthorTom Leinster
    • CommentTimeJul 28th 2011

    I've just started to go through it, and am very pleased at how efficiently Andrew's script has worked.

    Inevitably some tweaks are needed, though. Here's my list so far.

    1. Sectioning: the original version has an unnumbered, unheaded introductory section. This has now become a section labelled "1. Introduction", thus changing all the other section numbers. It's a slight disappointment that the Lab and ArXiv versions don't match in their section numbering, though it's not a big deal.

    2. Diagrams and symbols: this is more of a note to self. Not all of the arrows in the diagrams, or special symbols, are showing up, so I need to check it on another computer/browser.

    3. Punctuation after diagrams: some of my original diagrams finished with punctuation, which has vanished in the converted version. I'd like to put it back.

    4. Inconsistent use of font for terms being defined. E.g. the definition of characteristic function is in bold italics, whereas the definition of subobject is in bold upright. (In both cases the surrounding text is upright, and in both cases I used the macro \demph.) Another example is in Definition 5, where the two defined terms get typeset differently.

    5. The quote environment didn't convert properly. This is in the paragraph beginning "Now we show that this is equivalent to the original definition". In the converted text, the quoted passage gets rendered as ordinary prose.

    6. Examples 6: these are all set in italics, which is hard on the eyes. I'd prefer them to be upright, as in the original text. This was also done in Revision 22 (Jan 27).

    7. Examples 6: the numbering of the individual examples within Examples 6 looks funny to me. E.g. "1." is on a line all on its own, followed on the next line by the text of example 6.1.

    8. Examples 6: personally I'd prefer 6(i), 6(ii) etc rather than 6.1, 6.2 etc, but it's no big deal.

    I think I'll do this in small batches. I'll get going on these eight now.

    • CommentRowNumber134.
    • CommentAuthorTom Leinster
    • CommentTimeJul 28th 2011

    Hmm, some of these are posing a challenge.

    I don't understand what's behind problem (4). E.g. if you look at the source of Definition 5, you see the same formatting commands applied to the terms "topos" and "elementary topos": yet they render differently (for me at least).

    I don't think I know how to fix problems (6) or (7), though I guess I could figure it out.

    • CommentRowNumber135.
    • CommentAuthorTobyBartels
    • CommentTimeJul 28th 2011

    Sometimes using underscores (instead of asterisks) to indicate emphasis behaves strangely next to parentheses and other punctuation. (Somehow, this is supposed to be a feature of Markdown.) I’ve never understood this; also I have no idea what {: style="font-style: normal"} is for and how that might interact with this bug/feature.

    • CommentRowNumber136.
    • CommentAuthorUrs
    • CommentTimeJul 28th 2011

    Thanks to Andrew and Tom for their work! Apart from the small remaining bugs, it looks very good, I think.

    One feature seems to be missing, though: the hyperlinks. I think we should hyperlink each technical term to its nLab entry at least once when it appears first. I’d be even inclined to hyperlink each technical term each time it appears, so as to allow readers to jump right into the text in some paragraph while trusting to have the power of the nLab behind them to help them figure out technical meaning.

    • CommentRowNumber137.
    • CommentAuthorAndrew Stacey
    • CommentTimeJul 28th 2011

    Incidentally, the inconsistencies in formatting are as likely to be due to bugs in the CSS as bugs in the script. The Firefox plugin Firebug is essential for debugging this sort of thing. Other browsers have similar plugins (some have then as part of the actual browser).

    • CommentRowNumber138.
    • CommentAuthorAndrew Stacey
    • CommentTimeJul 28th 2011

    The inconsistent use of fonts is a line break thing. Maruku processes stuff line-by-line and there was a line break interrupting the extra font information in the curly braces. That’s now been fixed in my package (by setting the page width so large that unexpected line breaks are nigh-on impossible). See the Sandbox for an example with and without the line break.

    • CommentRowNumber139.
    • CommentAuthorAndrew Stacey
    • CommentTimeJul 28th 2011

    Okay, looking through the list:

    1. I’ve made it unnumbered in the script now, do you not want a section heading there at all? The issue is more about the table-of-contents than anything else.

    2. What browser are you using? (And do they turn up again in, say, Firefox?)

    3. Whoops. That’s probably on those diagrams where I’ve had to put in two versions: one for the web and one for paper. Would it be too much trouble for you to list them?

    4. This seems to be a line-break thing (see previous message) now fixed.

    5. No idea why this didn’t work. Should be fine in the next update.

    6. Agree. Are there any other “theorem like” environments that you’d like to be upright instead of italic?

    7. I don’t see this with FF. Which browser are you using? It’s probably a “feature” of the hack that I’m using to get labels on the items. If I can figure out a way to fix this, I will. (I’ve put in a Feature Request for a proper way of doing this, but until/if that gets implemented, I need to use some sort of hack here.)

    8. I’ll need to look up the CSS for this. I can certainly do the roman numbering, will have to look up how to do the parentheses.

    • CommentRowNumber140.
    • CommentAuthorAndrew Stacey
    • CommentTimeJul 28th 2011

    Regarding list syntax: do you want all lists to be roman, or just those in Example 6?

    • CommentRowNumber141.
    • CommentAuthorAndrew Stacey
    • CommentTimeJul 28th 2011

    Ah, I’ve just thought of a problem with that. When you refer to an element in a list, what is shown on the page is something like 6(4). The fact that these are numbers is chosen by maruku. I don’t know if it is possible to change that. So whilst it is possible to change the numbers in the list easily by CSS, changing how they are referred to is more problematic. I could do it, but I’d like to know whether or not I should.

    In a little more detail, it’s a chicken-and-egg thing. The number 4 appearing in the list is rendered by your browser and until your browser generates the page, no one really knows what will appear there. You could override this to produce a (picture of a) chicken if you liked. What maruku does is to count how many things are in the list and guess what will occur. So it says, “I’ve counted four things in the list so I’m going to guess that the marker will be 4.” Now I can make my LaTeX package do that guesswork instead and make label in to a iv, but the problem with that would be that if someone (say, you) came along later and said, “We should refer to item 4 elsewhere as well” then you’d put \ref{whateveritwas} in that place and maruku would then render it as 4.

    My policy with this package is: if maruku (itex, instiki) can do something, let maruku (itex, instiki) handle it as that gives the greatest flexibility with editing and later customisation. As I said, I can go against this principle, particularly given that these articles won’t be edited all that much, but I’d like to know how important it is first.

    (This isn’t a problem in ordinary LaTeX because then LaTeX is in charge of the whole process, including deciding how the labels will be displayed.)

    • CommentRowNumber142.
    • CommentAuthorTom Leinster
    • CommentTimeJul 28th 2011

    Thanks for the comments. Here's a further question: is it worth me changing things by hand, or are you going to run your automatic conversion process again? (I've changed a few things by hand so far, but not so many that it's important.)

    1. I see that all the section headings are unnumbered now. This isn't what I wanted, and I'm not sure you intended to do that. My ideal would be to have the introduction unnumbered, and the rest numbered. (I don't mind whether or not the introduction bears the heading "Introduction".) But I don't feel strongly about it. My second choice would be to have every section, including the introduction, numbered. My last choice would be to have no section numbered.

    2. I was using a very ancient version of Firefox, so let's ignore that. Right now I'm using an up-to-date version of Firefox. The diagrams render fine, though there are various symbols that don't appear: e.g. blackboard bold A and typewriter t (the truth values map 1 --> Omega). Doubtless there's something I should install.

    3. I fixed them by hand, but in case it's useful, here's a list. Page numbers refer to the latest arXiv version (v3). There are diagrams with punctuation afterwards on page 3 (three diagrams), page 4 (one diagram), and page 24 (one diagram).

    4. Thanks.

    5. Thanks.

    6. My personal preference is to have formal assertions of fact (theorems, propositions, lemmas, corollaries, conjectures, etc) in italics, and other things (examples, definitions, remarks, warnings, etc) in upright. Some people like definitions to be in italics too; I don't feel strongly about that.

    7. That was with the aforementioned very ancient version of Firefox, so again let's forget it. It looks fine with the up-to-date Firefox I'm using now.

    8. This sounds like too much trouble. Let's stick with arabic.

    • CommentRowNumber143.
    • CommentAuthorAndrew Stacey
    • CommentTimeJul 28th 2011
    1. That was a temporary bug in the script! I haven’t been updating the nLab copy every time I spot a minor bug because I can be pretty sure that there’ll be a major one along soon. They’re back as numbered, but for some reason my browser is choosing to number then as 0.1 and so forth.

    2. That’ll be the Ol’ Stix fonts. Get yourself a decent set of fonts and you’ll be fine.

    3. I think I found them and corrected them in the original. I’ll check against your list.

    4. (junk to keep the numbering the same)

    5. (more junk, doesn’t anyone clean up around here?)

    6. Right, I’ve had a go at this. Annoyingly, it’s not as easy as it looks because of CSS inheritance: the obvious change made all the maths go upright as well. I think I’ve fixed this, but if you spot anything that should be upright/italic but isn’t, please tell me.

    7. Good.

    8. Thanks!

    • CommentRowNumber144.
    • CommentAuthorAndrew Stacey
    • CommentTimeJul 28th 2011

    Okay, fixed the weird section numbering.

    There’s a couple of outstanding minor errors to do with lists that I know how to fix. Ignore them.

    What do you think of the style for the references? In particular, how should the tags be displayed (I mean things like the “Johnstone(2003)” which are at the start of the line)? Should the bullet points be there?

    What about the thanks? I just picked something at random for that: smaller text and slightly “boxed”. Any ideas?

    I think that there are a few equations that Domenico had improved in the original Leinster2010 that haven’t made their way across, as well.

    • CommentRowNumber145.
    • CommentAuthorUrs
    • CommentTimeJul 29th 2011
    • (edited Jul 29th 2011)

    At the very beginning of Leinster2010 I have changed the “is to be submitted to” to a “has been submitted to and accepted by” with a pointer to the referee report.

    Andrew and/or Tom, please give me sign when the content of that page is stable. I will then help adding hyperlinks and then announce the publication on the nnCafé.

    Meanwhile, I can say that there is going to be a further submission to the nnJournal in a short while (probably at least).

    • CommentRowNumber146.
    • CommentAuthorTom Leinster
    • CommentTimeAug 1st 2011

    Thanks for the edits, Andrew, and sorry for the hiatus.

    Earlier I asked (but I guess you missed it): is it worth me making changes by hand? Because I don't understand the technology you're using, I don't know whether you're repeatedly running your script/program/... on my Latex source, in a way that would erase any changes I'd made by hand. If you tell me that this isn't the case, then I can plunge in and change various little things. But if it is the case, then I guess I need to wait.

    Re references: I like the reference style, except that I'd prefer "Johnstone (2003)" to "Johnstone(2003)". That is, I'd prefer a space before the open parenthesis.

    The "thanks" look good to me.

    • CommentRowNumber147.
    • CommentAuthorAndrew Stacey
    • CommentTimeAug 1st 2011

    I didn’t miss it, but I thought I’d answered that by email. My apologies for the crossed thread!

    I’m working from your LaTeX source. I’ve had to make a few modifications to that, but in a way that doesn’t affect the normal LaTeX output (at least, that shouldn’t affect the LaTeX - I haven’t downloaded diagrams.sty so I can’t actually compile the original!). Basically, I’ve copied the beamer \mode syntax that lets me put mode-specific code in certain places since there are some things that need doing differently in LaTeX to the nLab (diagrams being about 99% of this). So no, don’t change the nLab (doriath) page as I overwrite it each time.

    However, do feel free to change the LaTeX source if there’s something there that needs changing and send me the new version, I can easily merge those changes. For example, the reason that there isn’t a space in the references is because there isn’t a space in the source (take a look at the bibliography at the end of your tex file). Now, the important question is: do you want the spaces also throughout the text? If you look in, say, the introduction then you will see that the references are referred to by that tag, “Johnstone(2003)”. If you want a space there as well then that’s easy to do: we simply put a space in the source file. However, if you want a different syntax in the text to at the end, then that is probably something to be done at the very end as it’s a bit special so not something that a generic script should handle.

    • CommentRowNumber148.
    • CommentAuthorTom Leinster
    • CommentTimeAug 1st 2011
    • (edited Aug 1st 2011)

    Thanks. I think I see the mail you're referring to, but I'm afraid I didn't understand the implications of what you wrote for my question about changing things by hand. Sorry.

    Regarding the bibliography, it was automatically generated by Bibtex (as I'm sure you're aware). I don't know what it does with the tags such as "Johnstone(2003)"; they don't appear anywhere in the text or in the bibliography. To answer your question, yes, I do want the space between "Johnstone" and "(2003)" in both the text and the bibliography.

    There are a couple of tiny changes to the text that need making (including a correction that I've OKed with Urs and the referee). I'll send you an edited version, and when I make the edits I'll insert the aforementioned spaces.

    • CommentRowNumber149.
    • CommentAuthorAndrew Stacey
    • CommentTimeAug 1st 2011

    I’ll wait for that version then.

    There’s one diagram that is still not quite right: the first example of “Examples 11”. The version in the Sandbox is right (as regards the arrows, I mean) so I need to look carefully to see why it doesn’t work in the article.

    Also, I need to put the diagrams in Examples 9 next to each other, otherwise the previous sentence doesn’t make sense.

    • CommentRowNumber150.
    • CommentAuthorTom Leinster
    • CommentTimeAug 2nd 2011

    "New" version (with aforementioned micro-edits) sent by email.

    • CommentRowNumber151.
    • CommentAuthorAndrew Stacey
    • CommentTimeAug 8th 2011

    Right, I think I have the references sorted out, and that tricky diagram (bug in the code). I hope that all the internal links work correctly: someone should go through and click on everything to check that.

    One minor wording query: in the acknowledgements you say that the diagrams were produced using the diagrams.sty package. That’s not true of the nLab version. But I don’t know if we want to go down the slippery slope of modifying the wording where it doesn’t make so much sense on a webpage …

    • CommentRowNumber152.
    • CommentAuthorTom Leinster
    • CommentTimeAug 8th 2011

    Excellent: thanks for the new version. I'll go through it and look for further issues.

    Personally, I think we should remove that sentence about diagrams.sty. I think this is a clear-cut case.

    • CommentRowNumber153.
    • CommentAuthorTom Leinster
    • CommentTimeAug 10th 2011

    Here are comments and corrections for the latest lab version. As before, page and line numbers refer to the most recent (v3) arXiv version.

    1. (Minor) Dashes in the latex (typed as ---) look as if they're being rendered as en-dashes rather than em-dashes. I like to use them without surrounding spaces, and I think the way that's coming out here looks wrong. Example: page 2, at the end of the paragraph next to the little diagram. I'd favour either making the dashes longer or putting spaces before and after.

    2. Acknowledgements: I'd like to remove the last sentence (re Paul Taylor's macros), as it wrongly implies that the diagrams on the web version were made with PT's package. I suspect PT himself would object to this.

    3. Page 5, lines -8 and -7: the displayed condition ("for every object..."), which in the original was done in the quote environment, is not displayed here or distinguished in any other way. I think this problem also came up earlier in the process.

    4. Page 7 (beginning of section 2): there's something wrong with the list formatting, at least as I see it (Firefox 3.6.18). The first item begins "1. 1. The terminal...", and similarly "2. 2." etc. for the other items.

    5. Page 7: in the aforementioned first item, there is a little diagram involving a parallel pair of arrows labelled $f$ and $g$. In the lab version, $f$ and $g$ have been interchanged. Here this doesn't change the meaning, but there are later instances of this problem that do. (I'll mention them explicitly.)

    6. Page 8, line 5 (lab version: just after Definition 7): the lab version says "Property 2.3 is, then, ..." This is confusing, as there's nothing labelled "Property 2.3". Perhaps this is related to issue (4).

    7. Page 11, near middle, the paragraph with all the bold terms: in the original, there is a parenthetical sentence "See Example 3.2(iii)...". In the lab version, this comes out as "See Example 9(4.3)...", which doesn't make much sense. The link isn't quite optimal, either, as it takes us (or me, anyway) to the beginning of Examples 9, not the particular Example referred to (namely, number 3 in that list).

    8. Page 11, near bottom, line just before the display: same problem as in (5).

    9. Page 13, just after the display "${x \in X | fx = gx}$": same problem as in (5).

    10. Page 14, line 1: in the lab version, the words "continuous functions" come out in italics, but should be upright.

    11. Page 14, line 4, "Examples (i) and (ii)": in the lab version, this comes out as "Examples 3.1 and 3.2", but should be "Examples 8.1 and 8.2" or "Examples 8(1) and 8(2)", or maybe even just "Examples 1 and 2".

    12. Page 14, just before the displayed diagrams: as you've pointed out before, "triangle on the left" doesn't make sense in the lab version. So either reformat the two diagrams or (surely easier) change to "such that the following triangle commutes".

    13. (Minor) Page 15, fourth display (involving a coend): towards the end of that display, there's a colimit. In the original, the text "$\rightarrow U, s$" is placed directly below the "$lim$", but in the lab version, it's below and to the right. I'd mildly prefer it to go directly below.

    14. Page 16, last display (in Defn 3.3): same as problem (5)... but this time it matters! It makes the definition of geometric morphism wrong.

    15. Page 18, just after Definition 3.7 (= Definition 14 in lab version): similar to problem (7). There are other instances of this; I won't list every one.

    16. Page 18, diagram: this looks wrong on my browser. I see this. Is this because I'm missing some fonts? If so, I'm slightly puzzled that I see an $a$ where there should be a vertical ellipsis. Usually when my browser encounters a character it can't display, it shows a square filled with tiny hexadecimal (?) symbols. Incidentally, you might have noticed that the diagram in the Latex version is now a little bit different from how it used to be, but don't worry about that; it's not worth changing.

    17. Page 21, bulleted list: in the lab version, the list ends at the wrong point. Some of the text that should follow the list has been absorbed into the second item.

    18. Page 24, third display (the parallel arrows labelled 0 and 1): the diagram in the lab version is wrong. (In an earlier incarnation of my paper, I had a pullback rather than an equalizer here.)

    19. Page 26, bottom: this is about marking the end of the "Aside". It's badly done in my original Latex, because although the end of the Aside is signalled with a \bigskip, that's made undetectable by the presence of a page break. Can we add some vertical space? The same goes for the end of the Digression on page 9. (There the same problem occurred, i.e. a \bigskip coincided with a page break. It's probably cosmic retribution for using \bigskip in the first place. I solved it, clumsily, by adding the words "Here ends the digression".)

    I feel bad writing out this long list, because I'm really grateful for all the work you've done, and almost everything has come out right. I hope this is making a useful contribution to debugging.

    • CommentRowNumber154.
    • CommentAuthorAndrew Stacey
    • CommentTimeAug 10th 2011

    I feel bad writing out this long list

    Don’t! That’s a fantastic bug report. I’ll take a look at those and see what’s what.

    • CommentRowNumber155.
    • CommentAuthorAndrew Stacey
    • CommentTimeAug 10th 2011

    (By the way, in case you’re not monitoring everything I do on the internet, I’ve already written three nLab pages and two blog posts using this package, so your bug reports are very, very useful to me in developing a system that I’m using quite productively.)

    • CommentRowNumber156.
    • CommentAuthorMike Shulman
    • CommentTimeAug 10th 2011

    I don’t know if we want to go down the slippery slope of modifying the wording where it doesn’t make so much sense on a webpage …

    It seems to me the “obvious” thing to do would be to have macros like \latexonly and \itexonly which are parsed or not parsed accordingly by (an appropriate style file for) LaTeX and by the latex-to-itex converter.

    • CommentRowNumber157.
    • CommentAuthorAndrew Stacey
    • CommentTimeAug 11th 2011

    It seems to me the “obvious” thing to do would be to have macros like \latexonly and \itexonly which are parsed or not parsed accordingly by (an appropriate style file for) LaTeX and by the latex-to-itex converter.

    Sounds like a good plan. Being a fan of beamer, I think I’d go for something like the beamer \mode command. Something like:

    \imode<latex>{%
    \newcommand{\parpair}[2]{\pile{\rTo^{\scriptstyle #1}\\ 
    \rTo_{\scriptstyle #2}}}
    }
    \imode<instiki>{%
    \newcommand{\parpair}[2]{%
    \underoverset{\quad #2 \quad}{#1}{\rightrightarrows}}
    }
    

    for that annoying \parpair command.

    (Incidentally, that’s what the problem with all those diagrams was: I can’t read. I got the arguments for \underoverset the wrong way around.)

    • CommentRowNumber158.
    • CommentAuthorAndrew Stacey
    • CommentTimeAug 11th 2011

    Right, here’s the first run through the list. I’ll not go through them item-by-item, but give an overview.

    1. en/em dashes. They are now the right dashes (they weren’t before). Whether or not you think that they are long enough is for you to decide. If you want more space, just say.

    2. The arrow labelling. As hinted in my previous comment, I had the arguments the wrong way around on the substitute command. Now fixed, so this should have fixed all incarnations of that error.

    3. Double labelling on the bold list. Fixed, it wasn’t a figment of your imagination (though it was a figment of your TeX! You had \item[\textbf{1.}] and so forth. There are better ways of achieving the same effect.)

    4. Diagrams: not all sorted, the mysterious a is due to a character-encoding issue and something I need to think about. I’ve put the other diagram alongside, but I need to tweak the vertical spacing.

    5. Internal references: I’ve chopped and changed how these have been done, so I think this is due to code confusion. I need to sort that out.

    6. Demarking special text (the “aside” and “digression”). I don’t think that vertical space is the right thing here, it would have to be a large space to make it obvious since paragraphs are already separated by space. There are various possibilities for what to do, perhaps the simplest would be to indent these sections. For an example of the sorts of things that are possible, take a look at topological notions of Frölicher spaces.

    • CommentRowNumber159.
    • CommentAuthorAndrew Stacey
    • CommentTimeAug 11th 2011

    Internal references are now sorted out (for the time being … I don’t think I’m doing it as “cleanly” as possible yet, but it works for this document and for my test document).

    • CommentRowNumber160.
    • CommentAuthorTom Leinster
    • CommentTimeAug 12th 2011

    Thanks very much!

    By my reckoning, the outstanding items from my list in comment 153 are numbers 13, 14, 16 and 19. To them I'll add:

     20. Page 16, equation (6): the adjunction symbol is upside-down. (That's maybe a misleading way to put it, because it's the same way up as in my original, but the directions and labels of the arrows have been swapped.)

    Regarding your comment 158: for item 3, I admit that that was a shabby hack. Guilty. I console myself that I am thereby being a typical author. Re item 6, I thought at first you were introducing me to the new verb "to Denmark", but closer inspection reveals that the truth is less exotic. Indentation seems a good solution; it looks good on your Frölicher page.

    • CommentRowNumber161.
    • CommentAuthorAndrew Stacey
    • CommentTimeAug 12th 2011

    Okey-dokey:

    • 14 and 20: This was the same underlying problem as (5): that I got the arguments to \underoverset the wrong way around. (I’m right, aren’t I, that these two are the adjunction diagrams that are sort-of identical - at least, from a typesetter’s point of view.)

    • 13: A possible bug in itex, or at least a place where it differs from LaTeX. The syntax \displaystyle \lim_{n \to 0} doesn’t work in itex as it does in LaTeX. I’ve told Jacques and put in a fix for the time-being (a fix that will work in both syntaxes so won’t need unfixing if Jacques changes itex). For those interested, here’s what it should show: lim n0\displaystyle { \lim_{n \to 0}} but it actually shows lim n0\displaystyle \lim_{n \to 0}. Somehow, the \displaystyle gets tangled up with the \lim preventing itex from seeing that it is a \lim and so treating the subscript as a regular subscript instead of something to be typeset underneath. The fix is an extra grouping: \displaystyle { \lim_{n \to 0} }.

    • 16: I’ve enabled unicode support in my package which fixes this. (The reason that you didn’t get the “unknown symbol” was because the piece of the chain that didn’t understand unicode was happening earlier.) I haven’t redrawn the diagram, though.

    • 19: I’ve indented the Aside and the Digression. What do you think?

    Still to do: Align the diagrams in Examples 9 correctly, redraw diagram on Grothendieck topologies (shouldn’t be hard to do: looks like I just need to remove two arrows).

    • CommentRowNumber162.
    • CommentAuthorAndrew Stacey
    • CommentTimeAug 12th 2011

    PS, regarding:

    I console myself that I am thereby being a typical author.

    and therefore being a perfect test-case for this process! Having written the package, it’s very easy for me to write documents that conform to its “standards”. Having to deal with someone else’s LaTeX means that I have to think about the right way to do things and figure out robust solutions. It’s been very useful.

    Regarding the bold list, may I recommend the enumitem package. It’s very versatile and allows you to customise the labels as you like.

    By the way, do you want that diagram of the dependencies of the sections drawn?

    • CommentRowNumber163.
    • CommentAuthorTom Leinster
    • CommentTimeAug 21st 2011

    Sorry for the delay in responding. Andrew, thanks for the further fixes. I like everything you've done.

    Regarding the diagram of dependencies, no, let's not worry about that. Similarly, don't worry about redrawing the diagram of a covering family (cf. your comment 161).

    I've just been through the whole thing again. Here are all the outstanding issues that I noticed.

    1. Alignment of diagrams in Examples 9 (cf. again your comment 161).

    2. Page 15: a small chunk of text has got deleted. It's the chunk "Second, every adjunction ... from the adjunction above as" just before the display involving "Sh(X)" and "Et(X)".

    3. Something's wrong in Examples 11. It should be a list containing two items (called (i) and (ii) in the original). As I see it (image), the first item is not correctly indented, and both items are labelled "1".

    4. This isn't really a problem, more a question or perhaps a feature request. Will \widehat, \widetilde etc. ever be supported? (Or are they now, in a way that I'm not aware of?) E.g. just after Definition 14 there's a widehat that's rendered as an ordinary hat, and looks just a little foolish, though I don't think it's important enough to be worth changing if it takes nontrivial effort.

    5. Page 21, middle: there should be a paragraph break between "... Set-enriched categories" and "How much has been lost ...", but in the lab version there isn't one.

    6. Theorem 16: the mathematical expression in the statement of the theorem should be in upright type, but appears in italics. The same is true in Theorem 17.

    7. Add a whole lot of links to nLab pages. I can go through and do it by hand, but was there a decision on a policy for this? I lost track of the discussion.

    • CommentRowNumber164.
    • CommentAuthorTobyBartels
    • CommentTimeAug 21st 2011

    I can address #4, since it’s come up before. The answer is that \widehat and \widetilde are already supported, by both iTeX and MathML, but (in my limited experience) I’ve never seen a browser correctly implement the resulting MathML.

    • CommentRowNumber165.
    • CommentAuthorAndrew Stacey
    • CommentTimeAug 22nd 2011

    In response:

    1. Still to do.
    2. Fixed. It was because this is the paragraph that you refer back to, so I put an “anchor” there, but because the anchor was actual XHTML, Maruku thought that the whole paragraph should be proper XHTML, but it isn’t. A newline fixed that.
    3. This and 5 are actually due to page breaks! TeX outputs a PDF which then gets converted to text. The PDF-to-text does a good job of preserving blank lines, but when a blank line occurs on a page break then it can’t tell that one should have been there. One solution is to make the page rather large. Something more robust would be to make blank lines non-blank and then remove them later (I already have to do this for indentation so it would be no great hardship.)
    4. See Toby’s comment.
    5. See 3
    6. Hmm, this is a CSS thing and is to do with how the italic text in the theorem main text is achieved. I don’t know if there’s a CSS specifier for “All text except maths”. I’ll investigate.
    7. The intention was that this be done initially by a script. But it needs to be done last, as it’ll break the relationship between the LaTeX source and the nLab page.
    • CommentRowNumber166.
    • CommentAuthorAndrew Stacey
    • CommentTimeAug 22nd 2011

    Now had a go at the alignment. Certainly better than it was, is it good enough?

    • CommentRowNumber167.
    • CommentAuthorZhen Lin
    • CommentTimeAug 22nd 2011
    • (edited Aug 22nd 2011)

    @Andrew: Regarding (6), you could nest CSS selectors, e.g. .num_theorem [mathvariant="normal"] *, but the problem is that the .num_theorem * selector will still override the font-style on any children of those. It's probably better to not use the * selector at all and instead let CSS inheritance do what it's supposed to do.

    • CommentRowNumber168.
    • CommentAuthorAndrew Stacey
    • CommentTimeAug 22nd 2011

    Yes, the * selector does seem fairly greedy in what it applies to! I’ve taken it out of the CSS, so that the relevant rule (which is actually .thplain) is now just:

    .thplain {
    font-style: italic;
    }
    

    and that fixes that, so far as I can see. There may be a good reason for using the *, I’ll check with Jacques. Thanks for the advice!

    • CommentRowNumber169.
    • CommentAuthorTom Leinster
    • CommentTimeAug 22nd 2011

    Andrew (166), yes, the alignment looks fine to me now. Thanks.

    Toby (164): thanks. This reminds me of Tom Körner's comment (some time ago now) that he knows of no university where the students learn Lebesgue integration in their first year of undergraduate studies, though he does know of one where they're taught it.

    So I think where we are now is that there are two things left on the "to do" list: numbers 6 and 7 in my comment 163.

    • CommentRowNumber170.
    • CommentAuthorAndrew Stacey
    • CommentTimeAug 22nd 2011

    I’ve implemented Zhen Lin’s fix on doriath (which is where the page is currently hosted) for the CSS so 6 is fixed, just waiting to see if taking the * breaks anything else.

    So that leaves 7 …

    • CommentRowNumber171.
    • CommentAuthorTom Leinster
    • CommentTimeAug 22nd 2011

    Great. I'll do another full read-through at some point, but I think it would make sense to leave that until the hyperlinking is done. Thanks.

    • CommentRowNumber172.
    • CommentAuthorAndrew Stacey
    • CommentTimeAug 22nd 2011

    He he! I wouldn’t call the first run on the script a resounding success! Partly because we have pages in the nLab called and and I (or at least, have redirects there). Anyway, you can see it at converting latex to itex with links (doriath), but I warn you that it’s not pretty!

    • CommentRowNumber173.
    • CommentAuthorTom Leinster
    • CommentTimeAug 22nd 2011

    Yikes! :-) I now want to insert one of those little gif animations of someone pulling a face, of the type that John Baez likes to use, but I'll spare you all.

    One thing I believe we didn't settle was whether the wikilinks should be visible. My vote is for no, if they're going to be very dense. My ideal situation would be "no by default", i.e. you can make them visible if you want: cf. comment 110 and thereabouts.

    Anyway, I suspect that's the least of the challenges you're facing.

    • CommentRowNumber174.
    • CommentAuthorUrs
    • CommentTimeAug 23rd 2011
    • (edited Aug 23rd 2011)

    One thing I believe we didn’t settle was whether the wikilinks should be visible.

    Right, for a monograph-like text like this (as opposed to a more encyclopedia-entry-like text) it would be good to have the links not be too visible. Of course if they are entirely invisible it becomes questionable if they will still serve any purpose.

    On my personal web I have at some point changed the css so that hyperlinked words appear only in a somewhat different shade of grey (a shade a greyish green, in my case). I came to like that: they don’t heavily disturb the impression of the text but are nevertheless recognizable. There is of course lots of flexibility in the choice of shade.

    • CommentRowNumber175.
    • CommentAuthorAndrew Stacey
    • CommentTimeAug 23rd 2011

    Back to the CSS on theorems: Jacques says that there was a reason for needing the * selector but he can’t remember what it was for. Something was getting the style inherited correctly. But given that the * affects the maths in ways we don’t want, it is probably better to figure out what the specific problem was and target it directly than use this general method.

    I shan’t change the nLab’s CSS, but I’ve changed doriath’s so that Tom’s article displays correctly. At some point we should think about changing the main nLab CSS, but maybe some testing is in order first.

    • CommentRowNumber176.
    • CommentAuthorAndrew Stacey
    • CommentTimeAug 23rd 2011

    Okay, I improved the script a fair bit and the result is at converting latex to itex with links (doriath). Clearly, wikilinks in section titles don’t work!

    At the moment, it requires a person to go through and decide on each replacement. The current options are:

    • accept the replacement
    • skip the replacement
    • ignore the name for the rest of the document (avoids having to check “and” every time!)

    I think I should add:

    • always accept: mark this name as one that can always be replaced (should be used with caution, though).

    I also think that being able to undo an “ignore” flag would be useful (I accidentally set it to ignore “pullback” which was annoying). It would be nice if it could automatically ignore things like mathematics, but that might be too complicated.

    Whatever the script ends up like, it’s always going to need some manual adjusting afterwards. Also, for some bizarre reason, the processing of links seems to switch on and off in chunks. That’ll need a little investigating.

    One thing that I noticed was just how many normal words have mathematical meanings!

    (I haven’t done anything with the appearance of these links yet; at the moment, that would be counter-productive as we’re not sure that the links are correct)

    • CommentRowNumber177.
    • CommentAuthorAndrew Stacey
    • CommentTimeAug 23rd 2011

    Okay, I know what triggers the wikilinks to not be processed for a bit. I’ll have to go through the document to see if what triggers it can safely be taken out.

    (In the meantime, I’ve reported it to Jacques.)

    • CommentRowNumber178.
    • CommentAuthorAndrew Stacey
    • CommentTimeAug 23rd 2011

    All triggers removed. converting latex to itex with links (doriath) is now a hyperactive hyperlinked document.

    • CommentRowNumber179.
    • CommentAuthorUrs
    • CommentTimeAug 23rd 2011

    So the wiki page we see at that link is now entirely machine generated form LaTeX code?

    That’s very impressive!

    • CommentRowNumber180.
    • CommentAuthorRodMcGuire
    • CommentTimeAug 23rd 2011
    • (edited Aug 23rd 2011)

    Andrew - As I understand it, your Processing Script requires the user to answer questions about how to deal with individual potential links.

    Have you considered saving these answers as an ’Answer Script’ so that you can automatically re-run the processing? This would allow things like editing the Answer Script if some answer was wrong rather than having to manually re-answer all the questions.

    Of course this could fail if the source .TeX is changed unless you have some very tricky external bookkeeping relating answers to something like relative positions within the .TeX file.

    One solution would be to use the answers to modify the .TeX so that it encodes them. For now, I see this as the only solution for a word like ’and’ where in general you want to initially state that it should not be linked to the logical operator AND, but the source does contain a few ’and’s that you do want to link. If the modified .TeX initially stated ’don’t link “and”’ then there would be no need for marking any of the following plain ’and’s however there would be the ability to explicitly mark the logical ’and’s.

    I would assume that for now dealing with the word ’set’ must be frustrating in that it can appear in phrases like ’let’s set X = Y’ and is so prevalent in its math sense that you might prefer not to link every occurrence. I seems like a prime candidate for generally being ignored but deserving explicit marking for when it should be linked.

    • CommentRowNumber181.
    • CommentAuthorTobyBartels
    • CommentTimeAug 23rd 2011

    In principle, if we want to link every mathematical term, then sometimes we do want to link “and”. Example; “Definition: A foo is a bar that is hot and cold.”. Maybe it’s more convenient to assume that readers understand basic logic, rather than decide which uses of “and” are mathematically significant and which aren’t. On the other hand, if (as has been suggested) we decide to make links absolutely invisible (along with a note up top inviting the reader to click on any unfamiliar term), then having spurious links all over the page could be harmless.

    Writing that fake definition also reminded me of the problems of adjectives. One reason for the naming convention of using nouns in place of adjectives is that adjectives get reused more. We not only don’t have a page complete, that’s not even a redirect, since we don’t know whether to send it to complete space or complete category. There was an idea to create disambiguation pages for adjectives, and one was made, but most adjectives won’t produce automatic links.

    A related issue has an example right near the top: while “category theory” produced the link category theory, “category theorists” only produced category. More redirects could fix that. (There’s also an incorrectly accepted link to sections three paragraphs down.)

    • CommentRowNumber182.
    • CommentAuthorAndrew Stacey
    • CommentTimeAug 23rd 2011

    Urs (179): Almost all. The wikilinks are put in afterwards by a separate script. Also, there were one or two minor modifications that I made afterwards which will be taken up in to the LaTeX package, but which were only spotted in the last stages.

    Rod (180): Having been through that page once deciding on all the links, I completely agree that it would be good to save them! One could save them in the original file (I define a command \wikilink that gets converted to wiklinks), but there are two problems: one is that macros might expand to page names which wouldn’t be detectable until the document was processed; the other is that page names might go over more than one line (this is, I’m sure, not a large problem). The script was hacked together last night and this morning, so I’m sure that there is lots of room for improvement!

    Toby (181): Yes, as I was going through I did notice a few places where there were missing redirects. I should have taken note of them. The script works from the longest match to the shortest, by the way. And “section” was probably the most annoying of the lot since I knew that Tom did talk about sections in the document so I couldn’t just ignore it!

    • CommentRowNumber183.
    • CommentAuthorTom Leinster
    • CommentTimeAug 24th 2011

    So, Andrew, I'm a bit unclear as to what my role is at the moment. Is there something I should be doing now, or shall I sit back, let you perform heroic deeds, and await further instructions?

    • CommentRowNumber184.
    • CommentAuthorAndrew Stacey
    • CommentTimeAug 24th 2011

    Tom, I’m not so clear either!

    What would be useful would be if you went through the version with all the wikilinks and looked for ones that were wrong (either missing, wrong link - as in Toby’s examples, or linked-but-shouldn’t-be). If the list isn’t too long, we can fix them by hand. If it’s long, we’ll need to think again.

    Another thing to think about is which links should be visible and which invisible. Clearly most should be invisible (or nearly invisible, we can experiment) but there may be some that should be definitely visible.

    I’m not sure of the best format for recording this information. You could print out the pages, scribble on them, scan them back in, and send them to me.

    • CommentRowNumber185.
    • CommentAuthorTom Leinster
    • CommentTimeSep 9th 2011

    I followed your last suggestion, Andrew: print-scribble-scan. The result is at http://www.maths.gla.ac.uk/~tl/iitt_links_scan.pdf. The key to my notation is on the first page.

    I was awestruck by two things: first, how well the automated process works, and second, how almost every mathematical word has an nLab entry. It's incredible.

    In the light of the second source of awe, I'm even more certain than I was that links should be invisible. We can simply put the global instruction at the beginning "almost every mathematical word is clickable". Otherwise it punishes the world's eyeballs.

  2. awesome! the ideal solution would be having a clicckable “hide links/show links” option somewhee on the page. the “hidden links” version would have invisible but working links. so it would be to the reader which version to use from time to time, switching back and forth between the two at his convenience.

    • CommentRowNumber187.
    • CommentAuthorAndrew Stacey
    • CommentTimeNov 1st 2011

    I’ve done the links in accordance with your print-scribble-scan (very easy to follow, by the way). There were only a couple of places where either I couldn’t work it out or I couldn’t do what you said. The main difficulty is making a hyperlink with formatting inside it. As far as I can tell, this can’t be done by wikilink syntax. That is, you can’t do [[*amazing* wikilink]] to produce amazing wikilink. We could do it via manual links, or just apply the link to one of the words.

    My “report” is at Leinster-wikilinks.pdf.

    I also went through and created redirects on a few nLab pages (such as subtoposes) where it seemed obvious that these should be there.

    • CommentRowNumber188.
    • CommentAuthorTom Leinster
    • CommentTimeNov 1st 2011

    Wonderful! Thanks very much, Andrew. I went through all of them, and as far as I can see, it's perfect.

    Because it's now some way up the thread, here's the link to the article.

    And here are some points arising.

    As this has taught me, there are various words that are common in mathematical prose but also have precise mathematical definitions. Some that I noticed are: point, class, structure, true, state. Unsurprisingly, if I write "to make the point more clearly, we state the following result", the software adds inappropriate links to the words point and state. I suppose it might be useful to have a list of words of this type.

    This has also taught me that not only am I unable to decide how many hyphens to put into "finite-limit-preserving functor", I am also unable to decide how it should be hyperlinked.

    As far as I'm concerned, this article is now in its final state. Many many thanks to all who helped.

    • CommentRowNumber189.
    • CommentAuthorUrs
    • CommentTimeNov 2nd 2011
    • (edited Nov 2nd 2011)

    here’s the link to the article.

    Wait, this points within “Doriath” to an entry titled “converting latex to itex with links”. But the finalized article needs to appear at Leinster2010.

    And what about making the links less visible?

    • CommentRowNumber190.
    • CommentAuthorTom Leinster
    • CommentTimeNov 2nd 2011
    • (edited Nov 2nd 2011)

    OK, then it needs copying over. The version at Leinster2010 is an older one without links. The Doriath version is the correct, most recent one.

    Re making the links less visible, I agree that this should be done. (I was thinking of that as a general decision, not intrinsically to do with my article.)

    • CommentRowNumber191.
    • CommentAuthorTom Leinster
    • CommentTimeNov 2nd 2011

    it needs copying over

    Done.

    • CommentRowNumber192.
    • CommentAuthorUrs
    • CommentTimeNov 2nd 2011
    • (edited Nov 2nd 2011)

    Okay, thanks.

    As soon as we are happy with the way this article looks, we should make a public announcement to the nnCafé, saying something about how we are in the process of setting up a journal-like thing to go along with the nnLab, and that we have a first “issue”.

    Are we at this point? Or should we first change the appearance of the links?

    I kept saying that I, personally, would make the links be a pale shade of some color, as on my personal web. But in either case I am not sure what CSS code we need to insert in the entry. I only know how to change it for an entire web.

    • CommentRowNumber193.
    • CommentAuthorTom Leinster
    • CommentTimeNov 2nd 2011

    I'd vote for changing the appearance of the links first. First impressions, and all that.

    It probably goes without saying that I don't know how to do this myself.

    Here are a couple of comments on the home page of the Publications of the nLab. One is that it says "Annals" somewhere, where it should say nPublications. (I'd edit it myself, but can't.) Another is that it should presumably have a link to the list of published articles - a very short list, at the moment, of rourse.

    • CommentRowNumber194.
    • CommentAuthorTom Leinster
    • CommentTimeNov 2nd 2011

    In case anyone else is looking for the previous discussion about visibility of links, as I was just now, it's mostly at 107-129 in this thread (and then 173 and 174).

    If there was a button allowing the user to toggle between visible and invisible links, that would be perfect. (I think it was Eric who first suggested this.) But maybe that's hard.

    Urs, when you refer to the colour scheme on your personal web, do you mean pages like this? To my eye, and on this monitor, that's seriously unpleasant to read. The green is very bright. Presumably it's that way because you like it that way, and of course I'm only giving my personal reaction. I suppose the lesson is that people's preferences on link colouring can be very different (as if we couldn't have guessed...).

    • CommentRowNumber195.
    • CommentAuthorUrs
    • CommentTimeNov 2nd 2011
    • (edited Nov 2nd 2011)

    To my eye, and on this monitor, that’s seriously unpleasant to read. The green is very bright. Presumably it’s that way because you like it that way,

    No, I want it very pale. On my system it’s pretty pale. I’ll try to play with the color values further.

    • CommentRowNumber196.
    • CommentAuthorUrs
    • CommentTimeNov 2nd 2011
    • (edited Nov 2nd 2011)

    Okay, I have moved Tom’s refereed and finalized article to the write-protected nnJournal page.

    I have also done the following

    • tried to harmonize the formatting of the author names in the references, and tried to add the missing links there. But still not perfect. Maybe somebody else can have another look at it.

    • renamed the ambient web into “journal”

    • tried at HomePage (journal) to consistently change “Publications of” and “Annals of” into “Journal of”. But check.

    • created 2011 (journal), meant to contain the list of publications 2011 (to be followed by similar pages, clearly)

    • CommentRowNumber197.
    • CommentAuthorUrs
    • CommentTimeNov 2nd 2011

    Wait, there is some problem. At Leinster2011 (journal) appears at various places “error messages” (or whatever it’s meant to be)

    wikichunkwordchunk .
    

    What’s going on?

    • CommentRowNumber198.
    • CommentAuthorUrs
    • CommentTimeNov 2nd 2011

    At HomePage (schreiber) I have made the link color much paler. Is it better now?

  3. There’s some unpleasant issue going on: I perfectly see the Doriah version, but I see the version at Leinster 2001 badly corrupted, with many “Could not include SVG pullback” and the like. Presumibly there are commands used that are only defined and accessible from the Doriah page.

    Other issues:

    to me it has not been immediate to see that published papers were under “Arxive”. maybe a more explicit “nJournal issues” or “Published papers” or the like could be better.

    it is not clear where to get the referee’s reports for Tom’s article. I think we should link them from the 2011 page, together with the “submitted on/accepted on” dates

    also, if Tom’s paper is open to community edits, we should feep there links to both a freezed version with the accetance date and to the editable version

  4. At HomePage (schreiber) I have made the link color much paler. Is it better now?

    looks nice to me