> > Email is an event driven push messaging bus.  For that reason,
> > email is good.  Having to poll places for data doesn't scale,
> > having something come in over the socket and send a kqueue event
> > telling the listening process to check out the data that's
> > blocking on the socket and waiting to be processed is good.
> > Having multiple lists is like having multiple ports.
> 
> Unless you're running your own mail server (which most people
> don't), you have to have a process polling an external location for
> data.  Perhaps it's better to have single process (your E-mail
> clinet) poll a single location (your POP3 or IMAP server), but
> polling at least one place for data is inevitable. (Using E-mail
> polling as the sole source for retrieving certain types of data is
> another matter.)

True enough.  Let me amend my analogy slightly: if you're running
your own mail server, it's like kqueue in that it will wake up the
process and notify it that there's data on the socket.  For the rest
of the world, they have to poll the socket, but the data is already
waiting at the socket in a queue and just needs to be picked up.

> The specific technology may be irrelevant if the data is available
> in a multitude of forms, and on demand (e.g. I shouldn't have to be
> at my home PC, or require access to my mail server). The larger
> issue of an announcement list is the effect on the use and
> maintenance of the RAA.  Already there have been announcements to
> rubynet-announce of items that have no entry in the RAA.  How can we
> best promote and maintain a single authoritative location for
> library information (including additions and updates)?

::shrug:: cave to the fact that there are multiple stock markets for a
reason?  ::shrug:: Single authoritative places are good for standards,
but not necessarily for implementations or use.  Could you imagine a
world with only one software vendor?  :) -sc

-- 
Sean Chittenden