[Top][All Lists]
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
[Pan-users] Re: Re: Marking Messages Read
From: |
Duncan |
Subject: |
[Pan-users] Re: Re: Marking Messages Read |
Date: |
Tue, 09 Nov 2004 01:58:29 -0700 |
User-agent: |
Pan/0.14.2.91 (As She Crawled Across the Table) |
Paul Trevethan posted <address@hidden>,
excerpted below, on Sun, 07 Nov 2004 20:00:21 +1100:
> BTW, is there some way to send my 'method of operation' to the
> developers as something for them to consider as options in future
> development work?
A question without an easy/direct answer.
Yes: This list/group provides one way of getting the visibility up. For
longer term tracking, however, bugzilla it as a "wishlist" item.
HOWEVER
No: Official PAN development is more or less stalled at present, with the
last "beta" release nearing a year ago now (2004.01), and development slow
for another six months previous to that. The details are in the archives
as I and others have repeated them a number of times, but basically,
Charles (lead developer) said his job and meat-space life were taking up
all the time he had, at present. No real hint as to how long that might
last.
It's not /all/ bad, however. PAN had reached basic maturity at its
current level, and a number of big feature additions were on hold pending
implementation of a back-end database library (SQLite has been the working
basis so far). That's not Charles' strong point, and he seemed hesitant
to really get into it. The de facto development freeze has allowed
several developers with a bit more experience in the database field to
start experimenting, and the necessary work is /slowly/ coming together.
It's quite possible that if development /does/ resume sometime next year,
a decently working solution will be ready to pretty much drop in place.
If that happens, having that database backend opens up so REALLY
interesting new possibilities, including fully automated multi-server
handling with the ability to set up single views of newsgroups that
display what's available on /all/ configured servers, then download parts
as available first from the cheaper servers (say ISP bundled), then the
expensive backup servers (measured service pay servers) if necessary, all
by just marking them for download and letting it rip, with what server
it's coming from being handled entirely transparently to the user, if
desired.
As well and perhaps more immediately practically, PAN currently has
serious scalability issues as it gets into the million overview per group
range. A decent part of this is due to its use of primarily GUI widgets
with a bit of database functionality appended, for processing FAR more
data than they were designed to efficiently handle. Getting that all in a
decent database and using "database like" memory and efficiency
optimizations, should allow PAN to scale up FAR further than it can
currently, at /least/ by an order of magnitude, so we're talking ~20
million overviews, instead of the current ~2 million. Further, since the
database management techniques used would allow "memory windowing" such
that data that must now ALL be handled in memory would be efficiently
stored on disk, with only a "memory window" onto it, so in theory, scaling
would then be linear far above that 20M overviews, so 50-200M overviews
would be more manageable than even three million overviews are now. Of
course, that's essential for practical multi-server merged handling, as
well, but there are a few groups where it'd be handy on its own.
Thus, in some ways this could be seen as a necessary pause -- like the
scheduled hold in the countdown clock before a NASA launch. OTOH, we
really don't know if Charles will ever get back to it, and it may be that
someone else will need to take over. I have the interest, but am quite
some time away from having the ability. Others with the ability don't
really seem interested, altho in Charles' absence as I said, there have
been a few (2-3) db folks working on the code, mostly privately so far,
but making progress on what /might/ be a pretty big rewrite of the
back-end.
In any case, it's quite likely that something will happen next year.
Either development will restart, very possibly in a /big/ way, or by this
time next year as PAN reaches about two years without serious work, it'll
likely begin to be dropped as unmaintained by the distributions and
interest will move elsewhere. Two years is a LONG time in the open source
world!
As I said, this has been covered before. Take a look at the list archives
for how the discussion has changed over the months as the freeze got
longer, or at the developer list archive for the action, such as it is,
there.
--
Duncan - List replies preferred. No HTML msgs.
"They that can give up essential liberty to obtain a little
temporary safety, deserve neither liberty nor safety." --
Benjamin Franklin