Fwd: Postgres forums ... take 2

On 16/11/2010, at 2:01 PM, Craig Ringer wrote:

> On 15/11/10 17:37, Thom Brown wrote:
>
>> That's actually some good work you've done there! I didn't know phpBB
>> supported bidirectional mailing list support.
>
> Yikes. Neither did I. I've always seen phpBB as the barren wasteland of
> web forums - forums full of half-page animated GIF signatures separating
> single lines of text, some kind of content-free zone of minimum
> information density. Maybe it can be configured to be better than that
> after all.
>
> How does it handle threading? Will forum threads be properly threaded?
> And will replies have the correct In-Reply-To: <msgid> header so they
> get threaded correctly?

It uses the message-id and in-reply-to header information, so a change in email subject will not effect the flow of the thread. However not all emails stick by those rules, a few emails do not carry the header information in which case the script strips the subject (to remove "Re:", "[GENERAL]") and then matches the topic by name as a fall back.

Email header information can be sketchy at the best of times, but this does a pretty good job at making sure almost all of the messages are handled correctly. In some rare cases when the in-reply-to is missing and the subject has changed it will create a new thread, but a forum moderator can click a button to merge the threads and all is fixed.

>
> Have you been in touch with the Pg list admins to make sure they're cool
> with this?

At this point its a good idea, who is the best person(s) to contact? I want to make sure anything I do does not in any way reflect badly on the community or seem like i'm doing anything dishonest.

>
>> A few points though. I think we'd need to disable smileys, bbcode, any
>> form of rich text formatting, flash or embedded images.
>
> Mostly agreed. Limiting signatures to 4/5 lines would be nice too.

All forum software limits the size of a signature to stop people abusing it. As the signature is a separate entity the mail that comes from the forum can contain or ignore the signature.

>
> Limited HTML is really useful on web forums, though, as it allows you to
> delineate code from other text. Unless the whole forum is set to
> monospaced text with preserved whitespace, that's necessary to ensure
> that code samples are readable.

That's one thing that can't be fixed when incoming emails are converted to forums posts the code blocks appear as normal text.

There is no use importing huge archives seeing as most of it is either too old to be relevant, bloat the forum and most people don't bother using a search anyway before posting. But it may be alright to say import the last few months of data (that include product release announcements, active bugs etc.) I have written another script which imports the mbox files just as a proof of concept:
http://forums.postgresql.com.au/viewforum.php?f=34
Those 7 threads were imported from a single months mbox file.

>
> --
> Craig Ringer
>
> Tech-related writing: http://soapyfrogs.blogspot.com/