[Mailman-Users] Efficient handling of cross-posting

[Dragon]

Mikhail Teterin wrote:
>Well, depending on the archival method, a message can appear in 
>both, with only a single copy of it being stored. This can be 
>achived, for example, using symbolic (or even hard) links. 
>Similarly, if a relational database is employed, the same message 
>can be referred to from multiple places. Hardlinks, for example, is 
>how one IMAP-server (cyrus, I believe) stores messages sent to 
>multiple recipients. But I was referring to search-results only. 
>Regardless of how the messages are stored, if I type the search 
>string and select several of the mailing list archives to search 
>through, the same message may appear in my search results more than 
>once. That duplication should not happen -- I don't think, anyone 
>would disagree.
---------------- End original message. ---------------------

I disagree. Completely.

I think you are trying to convince everyone your way is the only 
proper way to do it and I don't think you are seeing the wider issues here.

First, there is no search facility built into mailman. Any such 
functionality is due to use of a third-party search facility either 
integrated into mailman by whoever installed it or through a source 
such as Google's site indexing.

Second, if a message is stored in multiple archives then it SHOULD 
show up in the search results if you search those multiple archives. 
It should only appear once if you search only a single archive. If 
you use the default pipermail archiver or something like mhonarc to 
archive your list traffic, those separate hits in each list archive 
could indeed be different in that the thread links and previous post 
and next post links in them may (most likely will) point to different 
places as people reply to the message.

Doing anything else is, IMO, both unnecessarily complex and may serve 
to obscure the relationship of posts in a thread within a single list.

It would really be a nightmare to implement a multi-list-aware 
archiving system and search system. How do you keep track of what 
posts are stored where and which ones link to them in the forward and 
backward directions for each list (and those things ARE list 
dependent) if you only store a single copy? I know this can be 
implemented in a sophisticated RDBMS system but that requires a lot 
more complexity and having to install and run yet another software component.

Just remember, you didn't pay anything for the right to use the 
mailman software, it is freely available under the GPL and both 
written and maintained by volunteers. Feature requests are just that, 
requests, not mandates. The developers may choose to implement 
something or they may not. The fact that one of the core developers 
(Mark Sapiro) is telling you that his is not desirable should be an 
indication that the way it functions is probably not going to change. 
Which leaves you with basically two options, develop the changes you 
want yourself or live with the way it works.

Dragon

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
  Venimus, Saltavimus, Bibimus (et naribus canium capti sumus)
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~