Re: [Gnu-arch-users] Automatic archive discovery, take 1

[Aaron Bentley]

On Sat, 2003-12-13 at 23:43, Andrew Suffield wrote:
> Here's a rough-but-functional, lightweight archive discovery
> system. This is something of an experiment; if this works well, it
> could be rewritten as part of tla proper. First, a new hook for tla:

I agree with the idea of providing a way of autodiscovering archives,
and I appreciate the notion of making it extensible.

I also agree that namespace issues are always subject to the possiblity
of corruption/dispute.  Adding another layer of indirection cannot fix
this, only user intervention.
 
[snip]

> The list is processed in sequence; if multiple names could match, the
> first is used.

I'm not sure this is the correct approach.  If multiple matches occur
and they conflict, it might be wise to get some user interaction.  Could
--pick-first be an option (for scripts), with user intervention as the
default?

[snip]

> If you had a list of site-specific archives, you could do something
> like this in your archive-list:
> 
> address@hidden http://internal/archive-list
> * http://people.debian.org/~asuffield/big-archive-list

Can I suggest a third line?

address@hidden http://\1~\2/arch

What I mean is, for address@hidden, try http://example.com~/me/arch as a
last resort.  While this does require users to have web service using
the same hostname that they use for email, it's a very common practice. 
Those who have full access to their servers can map ~me however they
like, while those using ISPs will likely find it set up this way. People
not covered can register with you.

What I like about it is that it provides a way to locate arbitrary
archives without any need to register the archives with anyone.  The
namespace maintenance is ultimately taken care of by domain registrars,
and if you don't like their handling, you can always override it locally
(as with /etc/hosts).

Aaron

-- 
Aaron Bentley
Director of Technology
PanoMetrics, Inc.