Hi Jim,
On Sun, 2026-01-04 at 21:23 -0400, Jim wrote:
> On Sun, Jan 4, 2026 at 17:30 (-0700), Max Chernoff via ntg-context wrote:
> > On Sun, 2026-01-04 at 10:59 -0400, Jim wrote:
> > > Either would be good, although there might be things (speed, bandwidth,
> > > latency) to be said for having one on the west side of the pond. (I
> > > assume
> > > tug.org is hosted somewhere in North America.)
>
> > That would indeed be a reasonable assumption, but tug.org is actually
> > hosted in Germany.
>
> Huh. That's interesting (to me). (And thanks for gracefully saying my
> assumption was reasonable... my assumption was so strong I didn't bother
> doing a traceroute to see where the packets went. Duh!)
tug.org used to be hosted in the US (a decade+ ago), but they switched
to Hetzner (which is mostly Europe only) a while back since it was much
cheaper than the alternatives. But the most important role of the
tug.org server is retrieving the package updates from CTAN, repackaging
them for TeX Live, and then pushing the tlpkg files back to CTAN, and
since the primary CTAN server is in Germany, the location works fairly
well for that.
Nelson Beebe hosts a quasi-offical mirror in Utah though
https://tug.ctan.org/
https://ftp.math.utah.edu/pub/texlive/
https://ftp.math.utah.edu/pub/tex/historic/
so there's still a decent North American presence.
> One thing that just occurred to me... if the mirroring only happens in one
> direction, it wouldn't be good if people (absent-mindedly) updated a copy
> of the wiki when they really wanted to update the original, so I wonder how
> easy it would be for the mirrors to disallow the editing feature.
Probably the easiest solution would be to just have a fully-static site
that's automatically updated once a week or so by using something like
this
https://github.com/openzim/mwoffliner
> Alternatively, in principle updates could propagate from mirrors back to
> the original, but my gut tells me that opens up huge admin issues. (My gut
> is occasionally wrong (location of tug.org is a case in point), but even
> with my recent rash assumption, I'm doubling down and believing my gut
> again.)
I think that the only options to allow writes anywhere are for all the
systems to use the same database (bringing us back to a single point of
failure), to allow writes to different mirrors to get out of sync (which
seems like a terrible idea), or to set up some complex
clustering/sharding scheme (which sounds much too complicated). So I
think that having the mirrors be read-only is the only feasible
solution, which is probably for the best, since I'd personally much
rather just host a bunch of static pages than to run a database and a
bunch of other software.
Thanks,
-- Max
___________________________________________________________________________________
If your question is of interest to others as well, please add an entry to the
Wiki!
maillist : [email protected] /
https://mailman.ntg.nl/mailman3/lists/ntg-context.ntg.nl
webpage : https://www.pragma-ade.nl / https://context.aanhet.net (mirror)
archive : https://github.com/contextgarden/context
wiki : https://wiki.contextgarden.net
___________________________________________________________________________________
