Re: Architectural ideas

1On Thu, Feb 13, 2003 at 10:25:33PM +0100, Gerrit Holl wrote:
>Correct me if I'm wrong, but I think that for optimizing performance it
>is indeed good to have a few static index pages, because hard-disk space
>is far cheaper than starting a process over and over again (correct me
>if I'm wrong). So I think this 'might well be' should be said stronger,
>and should become 'will be'.

Not really; any practical platform for running web applications will
provide long-running processes.  

>When a message is sent to multiple lists, does it then count as one
>message? I don't think so, so each msg-id is unique, so the <listname>
>is not necessary.

If there's no <listname> in the URL, that means the message-id index
has to be for all the messages in the archive across all of the
lists. I'd like the ability to delete the indexes and regenerate from
scratch at any point, but requiring that all the lists be re-indexed
is unacceptable; imagine having to wait for all of python-list to be
processed because you wanted to update the catalog-sig.

Plus, if different lists have different settings (list A discards
messages with attachments, while list B keeps them), then the unique
identifier for the processed version of a message is really (listname,
message-id).

>What do we do with internationalization? 'Date', 'author' and 'subject'
>are, of course, email headers, but they don't need to be. For a Dutch

I don't see why URLs need to be i18n-able.  Generated HTML, certainly,
but not URLs themselves.  Maybe they should be /d, /s, /t, though,
instead of /date, /subject, /thread.

--amk                                                    (www.amk.ca)
Unhappy is he to whom the memories of childhood bring only fear and sadness.
      -- H.P. Lovecraft, "The Outsider"