# from Gabor Szabo
# on Tuesday 27 May 2008:
>> Yes, there's an easy fix to force a toolchain upgrade and solve the
>> supposed "bootstrapping" problem. No, nobody seems to think that
>> gutting the 02packages.details.txt.gz file is a good idea.
>
>I must have missed the part, but why do you dislike
>02packages.details.txt.gz so much ?
I don't have any dislike for it, and any technical failings it happens
to have are another topic. But, they might be a catalyst for moving
forward -- more on that below.
But, we've run around (and over and through) the "bootstrapping" problem
quite a few times and everything always comes back to the inability to
force an upgrade to an older CPAN/CPANPLUS client. A suggestion which
I've made in a few places is to deprecate and/or gut the index, namely:
the 02packages.details.txt.gz file.
We're probably not ready for it right this very moment, because the
clients still don't auto-upgrade (and they need should have some
facility for doing this safely (presumably with an *active* beta period
first), provide options to bypass the upgrade, etc -- and certainly
only forcing it when we reach a state where the old versions become
incapable of installing new code) but whenever we reach that point,
we'll still have some portion of users running *old* code. So, if
nothing can move forward without upgrading the client, it seems to me
that we might as well reflect that situation on the client side by
making it impossible to do anything with the woefully outdated client
*except* upgrade the client.
Taking into account two related points (from Adam and Schwern):
1. Desire to make big indices update faster.
2. Desire to have "stable", "testing", "unstable" views of the CPAN.
To accomplish either of those things, some changes to the index are
likely, so why not lay the groundwork by finally getting the clients
updated? This *might* play-out like:
1. Clients which are able to be remotely told that they must upgrade
are released, possibly with support for new index schemes.
2. New index schemes are deployed.
2b. Various peripheral tools are updated.
3. No new modules go in the old index.
4. A few months pass, some people upgrade, bugs are fixed...
5. The old index is stripped of everything which isn't a new CPAN or
CPANPLUS client.
At which point, the only thing you can do with an old client is upgrade
the client to one which knows when you must upgrade.
Or maybe we only get to #3 and the pain of being frozen in time is
obvious enough that those who want to move forward will upgrade.
This is all, of course, simply my current musing on a master plan.
Maybe there's an easier way. Or maybe by the time there is any sort of
refinement, buy-in, and completion of it, perhaps we'll have hit 2032
and everything will have been long-since upgraded, or possibly the
number of users still clinging to older tools will have dropped to ~6
and it won't be worth doing anyway.
Or, hey -- maybe the wxCPANPLUS client comes out at the end of Summer of
Code and everybody loves it so much that they upgrade daily.
--Eric
--
Minimum wage help gives you minimum service.
--David Schomer
---------------------------------------------------
http://scratchcomputing.com
---------------------------------------------------
