Bugs for Sorcery are reported here

[bugzilla-daemon AT bugs.sourcemage.org]

http://bugs.sourcemage.org/show_bug.cgi?id=10799





------- Additional Comments From jblosser-smgl AT firinn.org  2007-02-20 00:59 
-------
(In reply to comment #12)
> Is there some reason an
> http solution would be better -- note that some fallbacks are actually
> ftp-based... actually looking at this proposed solution, there's no ftp/http
> distinction anyways (fallback1.ftp.mirrors.sourcemage.org for example.  Some
> people might choose one protocol over the other for their network and may 
> want
> to explicitly tell sorcery to prefer one.

You just gave one answer to your own question. Explicitly having them all on 
the
same protocol was one understood and accepted limitation of the original
discussion around this, but if we just have a single 'fallback.sourcemage.org'
(with redundancy), it can host both http and ftp and then redirect to one of 
the
mirrors that is providing that service.  If it's just DNS, every single 
fallback
has to provide the same protocols, or sorcery has to track which has which, or
we get more complex with names like ftp.fallback1, http.fallback1, etc.  None 
of
which accounts for any of the problems with reliability of these lines vs dns
caching, etc.

> Though, andrew's is the only one that currently is ftp.  I can certainly 
> open up
> vsftpd to add another.

This is incidental to the real issues we have right now.  ibiblio is not
reliable enough to be our only source for data files.  Other people are 
offering
bandwidth, but not anything we can count on.  Since we opened this bug, 
ibiblio
has been down a few times, we lost Andrew's mirror from the list entirely 
(maybe
it's back now?), lost yours during your move, and even David's has been up and
down once or twice and is now very limited on bandwidth.  More people offer
mirrors but we've even less visibility on reliabilty on those and I don't want
us dealing with trying to keep this up all the time.  It should be automated,
real-time, and bulletproof.

We need something that leverages 1 or 2 reliable lines to point people at
whatever fat fallback pipes are up, and we need this for more than just the
source code fallbacks.  In the end using service redirects is a simpler 
solution
that should let us take advantage of a bunch of high-capacity, low reliability
mirrors without our users having to wait for timeouts/updates/etc.

-- 
Configure bugmail: http://bugs.sourcemage.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug, or are watching someone who is.