Re: [Gnu-arch-users] Arch Cache & cached archives

[James Blackwell]

>     > The problem I'm trying to solve is repeated downloads of the
>     > same data.

Tom Lord wrote:
> I see no reason to modify arch in the slightest little bit to solve
> *that* problem.  Surely it can be solved more cleanly with proxies.
> In many or most cases, simply using the existing mirror functionality
> more carefully is sufficient.  It is strange, imo, to contemplate
> modifying tla to solve the problem you stated there.

We've already got a cache, called "tla make-archive -M". With some minor
tweaks to Aaron's proposal, we can dump make-archive -M *entirely*.

This one is just a little bit smarter. Instead of being an explicit
cache (make-archive -M archivename-SOURCE), this would make in implicit
cache. Instead of having to screw with limits (as you do with pull based
mirrors), the cache automatically handles limiting. Instead of only
serving users that go through the effort of making a pull-based user, we
can serve *every* user.

With very minor tweaks (tla update-cache archivename), we can drop the
whole -M portion of make-archive entirely.

In simpler words, it affects the common case *very* positively in a 
lot of ways.

> A "smart cache" in arch would, as a side effect, very likely anyway,
> eliminate the need for an external solution (while accomplishing much
> else besides).  But that is just a happy accident, I think --- and one
> that we could useful give up on if it were convenient to do so.

I see it as a slightly less dumb cache than this:

tla register-archive -d archivename
tla register-archive archivename-SOURCE location
tla make-archive -M archive-name-SOURCE /some/archive/dir 
tla archive-mirror archive-name

> The way to build a "smart cache" is top-down, not bottom up.
> Meanwhile, tla shouldn't have a built in "dumb cache" when it so easy
> to construct one externally to tla.

If this logic were true, we'd be forced to remove archive-mirror -M
anyways, and tell users to rsync remote archives.

Anyways, I'm excited by this idea. I'm hoping we can get you onboard.


-- 
James Blackwell          Try something fun: For the next 24 hours, give
Smile more!              each person you meet a compliment!

GnuPG (ID 06357400) AAE4 8C76 58DA 5902 761D  247A 8A55 DA73 0635 7400