## Not signed in

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

## Site Tag Cloud

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

• CommentRowNumber1.
• CommentAuthorUrs
• CommentTimeSep 21st 2017
• (edited Sep 21st 2017)

As of recently the redirecting behaviour of the nLab software has changed, and it is breaking our conventions.

For instance

And interestingly:

etc.

I suppose it makes sense for us if accents are disregarded in entry identification, that could be useful. But we do need that lower case is distinguished from upper case, or else many existing entries will be hidden.

• CommentRowNumber2.
• CommentAuthorDavid_Corfield
• CommentTimeSep 21st 2017

I think I was right in my observation on the other thread that this affects only a few entries, set/Set, spectrum/Spectrum in addition to poset and topos.

You can see the point of such redirection for any case where a sentence begins with a linked term and so is capitalized. We haven’t typically added in capitalized versions for redirects. So if I say

Monomorphisms from one group to another…

it would be handy for this to go to ’monomorphisms’, plurals being items we do put in redirects.

But now I look at monomorphism I see it does have capitalized redirects. I’m sure we haven’t done this for many entries. Perhaps just as the occasion has arisen.

• CommentRowNumber3.
• CommentAuthorDavid_Corfield
• CommentTimeSep 21st 2017
• (edited Sep 21st 2017)

Now I’m very confused because I’m seeing ’Bad gateway’ for Monomorphisms which is there in the redirects of monomorphism.

• CommentRowNumber4.
• CommentAuthorUrs
• CommentTimeSep 21st 2017

I think I was right in my observation on the other thread

Sorry, I didn’t see that.

• CommentRowNumber5.
• CommentAuthorTodd_Trimble
• CommentTimeSep 21st 2017

By the way, we also have Pos. I might recommend a redirect of Poset to Pos; I just added a little to Properties at Pos.

• CommentRowNumber6.
• CommentTimeSep 21st 2017

This should be fixed now. Apparently it was a side effect of changing the database to utf8 encoding, as the default collation utf8_general_ci is for whatever reason case insensitive.

• CommentRowNumber7.
• CommentAuthorUrs
• CommentTimeSep 21st 2017

One thought: the insensitivity to accents might be useful to keep. Is it technically possible for entries to be case sensitive but insensitive to accents?

• CommentRowNumber8.
• CommentTimeSep 23rd 2017

Sure, done. Example: HöméPàgê.

• CommentRowNumber9.
• CommentAuthorDavidRoberts
• CommentTimeSep 23rd 2017

Test: homepage. Hmm still redirects…

• CommentRowNumber10.
• CommentAuthorMike Shulman
• CommentTimeSep 24th 2017

If you look at the source, “homepage” is an explicit redirect to HomePage, so that’s correct. The important ones to test are set and Set, topos and Topos, poset and Poset

• CommentRowNumber11.
• CommentAuthorDavid_Corfield
• CommentTimeSep 24th 2017

…and spectrum and Spectrum.

• CommentRowNumber12.
• CommentAuthorDavidRoberts
• CommentTimeSep 24th 2017

OK, didn’t think of that. Thanks.

• CommentRowNumber13.
• CommentAuthorUrs
• CommentTimeSep 24th 2017

• CommentRowNumber14.
• CommentAuthorTobyBartels
• CommentTimeOct 1st 2017
• (edited Oct 1st 2017)

Sometimes we've actually used accents in the names of pages (although this violates the naming conventions, it's a relatively harmless violation as long as the nonaccented redirect is made, or it least it was harmless until now), such as Kurt Gödel. So it would be nice if Kurt Godel went there automatically (it goes there, but only because it's an explicit redirect) or even Kürt Godel and Kürt Gödel (but they don't).

This requires normalizing all of the page names before searching for a match, not just normalizing the name of the page requested, so may not be easy to do. But perhaps it's enough to normalize the name of the page requested and then look for a match among explicit redirects (which would make Kürt Godel and Kürt Gödel work as soon as Kurt Godel does).

• CommentRowNumber15.
• CommentAuthorDavid_Corfield
• CommentTimeOct 1st 2017

So it would be nice if Kurt Godel went there automatically (it goes there, but only because it's an explicit redirect) or even Kürt Godel and Kürt Gödel (but they don't).

I think I’m missing something. Why should we want there to be misspellings?

• CommentRowNumber16.
• CommentAuthorTobyBartels
• CommentTimeOct 1st 2017

Let me clarify: I wouldn't want anybody to put any effort into making such misspellings as Kürt Godel or Kürt Gödel work. However, I would like to make both Kurt Godel and Kurt Gödel redirect to whichever of the other exists. (Currently, only one of those would work automatically, and it's not the one that we actually need.) If it's possible to modify the automatic redirection so that it would work either way, then we might get the misspellings to redirect for free, and that would be a small bonus. (But if the automatic redirection is modified in such a way that they don't come for free, then never mind.)

Note that the current automatic redirection algorithm does give us misspellings for free when the target is Kurt Godel, just not when the target is Kurt Gödel. (Compare the example of HöméPàgê in Adeel's comment #8.) So if automatic redirection can be modified to automatically redirect Kurt Godel to Kurt Gödel (not just the other way around), then it seems likely that the misspellings would come for free then as well. But if they don't, then that's fine too.

This was all supposed to be implied by the phrase ‘or even’. :–)

• CommentRowNumber17.
• CommentAuthorUrs
• CommentTimeOct 2nd 2017
• (edited Oct 2nd 2017)

I agree with Toby, it would be good if calling a page were accent-insensitive. Especially for names (category:people entries) this should be useful.

But now I see that the new redirecting behaviour only knows to strip off accents, but does not know to try to add them. That’s maybe no harm, but not as useful I I thought (in #7) it would be.

• CommentRowNumber18.
• CommentAuthorDavid_Corfield
• CommentTimeOct 2nd 2017

OK, so long as we don’t have an outbreak of superfluous accents. I’d rather it was indicated to someone writing Kürt Godel that there might be a problem.

• CommentRowNumber19.
• CommentAuthorUrs
• CommentTimeOct 2nd 2017

The issue is with names like José Gracia-Bondia. Even after some searching, I wasn’t sure what the exact precise decoration of the letters is. (Is it a plain “c”, a plain “a”?) Moreover, the spelling of the name is different even through different internet occurences of this author. It would be helpful if in creating an entry for this name, we would not have to second-guess all the possibly many combinations of leaving out/including these accents or other decorations.

• CommentRowNumber20.
• CommentAuthorMike Shulman
• CommentTimeOct 2nd 2017

It sounds like if we stick to the naming convention of leaving all accents out of the actual page name, then the automatic redirection will solve all our problems, no? In particular, we could just move Kurt Gödel to Kurt Godel.

• CommentRowNumber21.
• CommentAuthorTim_Porter
• CommentTimeOct 2nd 2017
• (edited Oct 2nd 2017)

Perhaps a simple (but less than perfect) solution would be to use the unaccented (therefore incorrect) form as the entry name but then to give the correct accented name in the first line of the entry. The page on Kurt Gödel would be called (for the purposes of the system) just Kurt Godel, but as now his name with accents would be the top heading of the entry, in those case in which the redirect mechanism is playing up. I notice that other accented names do not seem to cause such problems.

• CommentRowNumber22.
• CommentAuthorUrs
• CommentTimeOct 2nd 2017
• (edited Oct 2nd 2017)

It sounds like if we stick to the naming convention of leaving all accents out of the actual page name, then the automatic redirection will solve all our problems, no? In particular, we could just move Kurt Gödel to Kurt Godel.

I would like to put the correct decoration when known. The reason is that in most cases, certainly in this one, the symbol with or without decoration is regarded as a genuinely different letter by native readers, and to their eyes the wrong omission looks awkward. I think this is what David meant in #15.

If we don’t want to look foolish to native speakers, we should avoid writing “Godel” for “Gödel”.

My concern is just in cases where we are hard-pressed to know or find the correct decorations. In that case it would be good if the software was aware of the issue, for robustness of linking.

• CommentRowNumber23.
• CommentAuthorMike Shulman
• CommentTimeOct 2nd 2017

But we are writing in English, and in other cases when writing about people with foreign names in English we transliterate their names into the English alphabet. For instance, when writing in English about Владимир Воеводский, I don’t write “Владимир Воеводский” but rather “Vladimir Voevodsky”, because Russian letters like 'л’ don’t exist in English, and other Russian letters like 'В’ look exactly like some English letter 'B’ but actually mean something different. If 'ö’ is a different letter from 'o’, then it’s a letter that doesn’t exist in the English alphabet, so it seems to me that it makes perfect sense to transliterate it into its closest English equivalent when we are writing in English, even though all the other letters in the name do happen to transliterate to English letters that look identical. (I guess maybe Goedel is better to write than Godel, though.)

• CommentRowNumber24.
• CommentAuthorMike Shulman
• CommentTimeOct 2nd 2017

I guess maybe Goedel is better to write than Godel, though.

Of course, if we name the page Kurt Goedel then both the redirects have to be explicit in that case. But if we do this in line with a general policy of using English transliterations of all names, then we can consistently name the other page Jose Gracia-Bondia with no accents and let the automatic redirects solve that problem.

• CommentRowNumber25.
• CommentAuthorUrs
• CommentTimeOct 2nd 2017
• (edited Oct 2nd 2017)

True, for German Umlaute the well-established hack “ö” $\leftrightarrow$ “oe” is fine.

I am not sure how strongly people in other languages feel about the removal of their established symbol meaning. I seem to remember that some native Slavic or Russian speakers here on the forum did emphathetically object to some transliterations we used (I seem to remember Dmitri Pavlov raised such an issue once).

• CommentRowNumber26.
• CommentAuthorDavid_Corfield
• CommentTimeOct 2nd 2017
• (edited Oct 2nd 2017)

I should think we should try to get at least French and German names right. Now they’re readily available, umlauts are surely preferable to “oe”, “ae”, etc.

Voevodsky published under that name, and many Russian mathematicians of recent years have done likewise.

I’ve just done my bit for the evening and added acutes to “Poincare” in the titles of three pages. I see also there’s a Poincare symmetry transformations which is vacuous. Is that a page worth keeping after additions?

• CommentRowNumber27.
• CommentAuthorDmitri Pavlov
• CommentTimeOct 2nd 2017
I suggest that nLab employs the same transliteration as used by the mathematician himself in his publications.
Thus: Vladimir Voevodsky, Israel Gelfand, Mark Neumark, Paul Alexandroff, etc.
Of course, there is no reason not to add the untransliterated version as a redirect.
• CommentRowNumber28.
• CommentAuthorTobyBartels
• CommentTimeDec 6th 2017

Coming back to this late, here is some language trivia.

True, for German Umlaute the well-established hack “ö” ↔️ “oe” is fine.

This is more than a hack; ‘oe’ was the original spelling in German, and ‘ö’ developed from this in late mediaeval shorthand. See en.Wikipedia/Umlaut_(diacritic)#Umlaut#History.