Re: Gnus vs Wanderlust

[Jesper Harder <harder@ifa.au.dk>]

Miles Bader <miles@gnu.org> writes:

> colin.rafferty@morganstanley.com writes:
>> 1. It does not cache summary layout information.  When it builds a
>>    summary, it has to start from scratch.
>
> Is this actually a problem?  

Yes, from my benchmarking it is.

> Judging from what I see, the great bulk of the time is _not_ in
> summary generation, but rather getting info from the server

That may well be the case if you're fetching from an external NNTP
server (it also depends on network speed, of course).  But if you're
fetching from nnml or a local server like Leafnode, summary generation
is the major bottleneck.

> [My computer's pretty slow by contemporary standards too -- a 450MHz
> pentium]

I can beat that: 220MHz :-(

> I suppose one could try to cache and incrementally update summaries,
> but summary-generation is already so fast, even for very large
> groups, that it might make more sense to investigate whether it
> could be made even faster (maybe even with some new emacs
> primitives).

Doing some low-level caching would should be quite easy by memoizing.
I'm thinking about stuff like `gnus-extract-address-components' which
is likely to be called on identical strings many times during summary
generation.

It probably won't buy us any earth-shattering speedups, but maybe 2-3%
here and there.

-- 
Jesper Harder                                <http://purl.org/harder/>