Want to take part in these discussions? Sign in if you have an account, or apply for one below
Vanilla 1.1.10 is a product of Lussumo. More Information: Documentation, Community Support.
Lurie’s Kerodon is live.
I hope that Lurie or somebody helping him is parsing his existing LaTeX source for e.g. the higher topos theory book to get into the required format, rather than writing everything out by hand again. The latter would seem to be a needless and colossal waste of time.
Ideally if it is retyped, it is done even better than HTT, taking the advances of the past decade into account, like the Elephant taking on all the improvements in the literature to get The Book proof, as far as possible.
Here’s Bielmans: https://chat.stackexchange.com/transcript/message/47209266#47209266
It’s supposed to be a rewrite / restructuring of HTT / HA / SAG (or maybe some subset thereof), as far as I know.
https://chat.stackexchange.com/transcript/message/47209636#47209636
No, not all. The input is a TeX file and a tags file, and it spits out (the database for) the website.
So not a complete re-typing, but an update nonetheless.
Interesting! A rewrite sounds strange. A reorganisation makes more sense. But it looks to me like there must be a fair bit of labour in creating tags, etc, and it seems like Lurie might be writing that source himself; that certainly doesn’t make sense to me, it should be possible to first parse the existing source to get the tags, etc, and then tweak things as desired after that. But who knows, maybe Lurie himself does have some scripts to do something like that.
ad 3: https://pbelmans.ncag.info/blog/2018/10/17/kerodon and some explanations at https://gerby-project.github.io mentioning Lurie.
I saw some incoming visits from this page, so let me answer some things here too.
I will update the website for Gerby soon, and include a minimal working example for people interested in using the software.
Thanks very much for joining in!
The main thing I was expressing some concern about was a situation in which Jacob Lurie sees the benefits of an online textbook and is committed to making it work, but then has to labour to fit his work into the format required by a particular piece of software. Software comes and goes rapidly, and it would be rewriting by hand of the LaTeX source into a particular format that I would regard as likely to be a ’colossal and needless waste of time’. If there are mathematical/expositional reasons for a rewrite, that’s all well and good, of course, that’s Jacob Lurie’s decision to make. But I do hope that the input into the Gerby software is such that (e.g. is just LaTeX, perhaps with some simple additional bells and whistles to indicate where hyperlinks are desirable) it could be easily extracted into some other format in the future.
It sounds like you have these technological matters in hand, though, so that’s great.
May I ask where the website is being hosted, and who is paying for the server? If it is not impertinent, may I also ask if you are being paid for your work, or if it is a voluntary effort?
Maybe a few technological questions as well… I presume you have nginx or similar in front of the flask web server (as well as the usual python intermediary between the two)? SQLite seems a little bit of a strange choice to me, especially if you storing HTML in it, but maybe it will be OK, at least to begin with, for a mathematics textbook. Have you tried load testing it?
The only input is a TeX file, and the tags file. No need to modify the TeX source whatsoever as long as you label everything that gets assigned a number (well, we need to remove some packages from the preamble before processing the TeX in Gerby, but that’s it), although of course we have only tested things on the Stacks project and Kerodon. In theory we could just take HTT and put that online verbatim (early testing was done on HTT without any issues). Similarly, anyone with a TeX file that has labels everywhere can generate a tags file and put it online.
Okay, that last sentence was a lie, at least for now. The number of steps to get Gerby running is a bit too high to make it user-friendly, but it’s certainly possible.
The host is PythonAnywhere, they had the right combination of Python and SQLite that we need. I don’t know who pays, but the cost is absolutely minimal. I’m a volunteer, as long as it doesn’t interfere with research too much it’s all fun.
We could store HTML outside SQLite, but the amount of HTML per tag isn’t too big, everything is split up, we are not treating the DBMS as a file storage for large files. There’s also quite a bit of metadata involved. I know SQLite isn’t the quickest DBMS in the world, but we access it through a database abstraction layer, so replacing it would be easy if the need arises. It also makes it easier for people to get started with Gerby, no need to configure an actual DBMS etc.
The load testing is happening continuously at the Stacks project :). Which is significantly bigger than Kerodon, and running smoothly. If I remember correctly, the processing times for the Stacks project were at most like 0.5s for the most difficult pages (something like https://stacks.math.columbia.edu/tag/00AO for instance). For Kerodon the processing times are generally much below 0.1s, with a few outliers above that for some pages, that I can probably optimise now that I think of it.
Thanks for your reply! It’s great that these kind of projects are being explored, and that people like you are giving of their time to help run them :-)!
May I ask what the traffic levels are for the Stacks project (e.g. how many hits per second typically)?
I’m not quite sure what AWStats calls a hit and what it calls a pageview, but it’s either 1 view per 5 seconds or 1 per 20 seconds on average. Maybe I should do a more detailed analysis at some point. There’s no load issue whatsoever in any case.
Thanks! Best of luck with the project!
1 to 12 of 12