Re: [Context] [NTG-context] updates without rsync

[patrick - only (2) is for you] On Fri, May 29, 2009 at 18:07, Diego Depaoli wrote:

2009/5/29 Peter Münster :

...

If this is not possible, perhaps the firewall is open for http? Then someone else with a Windows PC could prepare for you an archive and put it on an http-server? Why not do it directly on minimals.contextgarden.net?

[OT - for binaries maintainer] Even sources should be packaged in a zip file. After recent Mojca's changes I rsynced for a whiiiiiiiiiiiiiiiiiiiiiiiiiiiiile [OT - for binaries maintainer] Even sources should be packaged in a zip file. After recent Mojca's changes I rsynced for a whiiiiiiiiiiiiiiiiiiiiiiiiiiiiile

1.) packaging sources as zip files would mean having to download everything every time when the most tiny bit changes. 2.) There used to be problems with download speed on Taco's server (I don't know if that's still the case). But now that there's some more space on garden, we could put those file on the garden instead if that would help. It seems to be only 300 MB, so that's minor in comparison to all other stuff that eats space on the server. (Patrick: I forgot if I replied you, but you definitely may remove the old copy from the old server.) 3.) We take svn version of XeTeX and pdfTeX. There are usually no problems with those as only a single file changes. But LuaTeX and metapost come from "tags". Whenever Taco makes a new "tag", svn has a brilliant feature to "copy the files" and all the files get a recent timestamp. Consequently all the timestamps of all files change (isn't that nice?) every time when a new luatex version is released. 4.) This update was an exceptional case. Maybe I did a mistake by not issuing "svn change repositoryname" (the syntax is different, don't know it by heart), but by deleting and fetching all the files from scratch, just to make sure that everything still works fine. But I'm almost sure that all the timestamps got changed. 5.) I added a -c switch to fetching the updates. This makes updating (negotiating which files to fetch) slower, but skips updating files whose timestamps differ (and not contents). It calculates the checksum of every file before asking server what to fetch. I'll leave that switch there for a while to see what difference it makes. Mojca