On Tue, Jan 12, 2010 at 10:24 PM, Tom Lane <tgl@sss.pgh.pa.us> wrote:
> "Joshua D. Drake" <jd@commandprompt.com> writes:
>> On Tue, 2010-01-12 at 10:24 +0530, Dave Page wrote:
>>> So just to put this into perspective and give anyone paying attention
>>> an idea of the pain that lies ahead should they decide to work on
>>> this:
>>>
>>> - We need to import the old archives (of which there are hundreds of
>>> thousands of messages, the first few years of which have, umm, minimal
>>> headers.
>>> - We need to generate thread indexes
>>> - We need to re-generate the original URLs for backwards compatibility
>>>
>>> Now there's encouragement :-)
>
>> Or, we just leave the current infrastructure in place and use a new one
>> for all new messages going forward. We shouldn't limit our ability to
>> have a decent system due to decisions of the past.
>
> -1. What's the point of having archives? IMO the mailing list archives
> are nearly as critical a piece of the project infrastructure as the CVS
> repository. We've already established that moving to a new SCM that
> fails to preserve the CVS history wouldn't be acceptable. I hardly
> think that the bar is any lower for mailing list archives.
>
> Now I think we could possibly skip the requirement suggested above for
> URL compatibility, if we just leave the old archives on-line so that
> those URLs all still resolve. But if we can't load all the old messages
> into the new infrastructure, it'll basically be useless for searching
> purposes.
>
> (Hmm, re-reading what you said, maybe we are suggesting the same thing,
> but it's not clear. Anyway my point is that Dave's first two
> requirements are real. Only the third might not be.)
The third actually isn't actually that hard to do in theory. The
message numbers are basically the zero-based position in the mbox
file, and the rest of the URL is obvious.
--
Dave Page
EnterpriseDB UK: http://www.enterprisedb.com