Thunderbird mailing-list / NNTP decision rationale

Andrew Sutherland asutherland at asutherland.org
Thu Aug 12 04:53:50 UTC 2010


  On 08/09/2010 08:05 PM, JoeS wrote:
> On 8/9/2010 10:00 PM, Andrew Sutherland wrote:
>> If something happened where our options were between sinking a 
>> non-trivial amount of MoMo man-hours into NNTP or removing the 
>> current C++ implementation, I believe we would remove it.
> Well, that "something" hasn't happened, and NTTP being a long 
> established protocol, that hasn't changed in years, that event is very 
> unlikely
> to happen. So why mention it as if it is a drain on resources.

Indeed, none of the changes we have made so far have affected the 
back-end or front-end in such a way that major refactoring work would 
have been required to the NNTP implementation to keep it working.

But that is not to say it has not slowed us down.  I know I spent a 
serious amount of time trying to get additional arguments for the NNTP 
message streaming process through the convoluted URI/service mechanism 
so that gloda could index newsgroup messages.  I did manage to get the 
information through, but I also apparently managed to regress handling 
of newsgroup attachments.  This was not an obvious thing because of the 
lack of test coverage.  I believe I've also spent numerous hours 
investigating crashers and pathological behavior that required 
consideration of the news codebase.  I have observed other developers 
spending time on issues specific to the NNTP implementation, as well.

The reason such things are worth mentioning is because we are now 
beginning to make back-end changes and front-end changes that could 
force the issue and because there seemed to be the impression that many 
of the Thunderbird developers have a problem with NNTP itself.  Because 
the mailnews database bits are reasonably abstracted from the protocol 
bits, I'm not expecting bienvenu's back-end changes to be that event, 
but the option needs to be on the table.  Likewise, the current 3-pane 
UI implementation will probably outlive us all, but a lot of the new UI 
experiments are gloda-centric and so the current news implementation 
will be invisible to them.  (And making the current news implementation 
visible to gloda is not a small undertaking; most of the work in gloda 
was in dealing with all the complexities of the POP and IMAP 
implementations.)

Andrew
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.mozilla.org/pipermail/tb-planning/attachments/20100811/ee5026cf/attachment.html>


More information about the tb-planning mailing list