Subject: Re: emacs buffer size limits (was: mozilla very unstable on 2.0rc4)
To: Matthias Buelow <mkb@mukappabeta.de>
From: Greg A. Woods <woods@weird.com>
List: netbsd-users
Date: 11/09/2004 16:17:10
[ On Tuesday, November 9, 2004 at 03:11:07 (+0100), Matthias Buelow wrote: ]
> Subject: Re: mozilla very unstable on 2.0rc4
>
> Greg A. Woods wrote:
> 
> > (I actually handle and answer all my mail using Emacs VM :-)
> 
> does VM somehow work around the 128mb buffer size limit in emacs these 
> days?  or do you use xemacs for larger mailboxes?

No, VM cannot work around the buffer size limit in any emacs (well not
without manually splitting a folder into multiple files ;-)

The biggest mail folder I have is 228MB of many copies of various e-mail
viruses and worms, but I've never tried to load it with VM.  :-)

(the next largest mailbox I have is just over 21MB and I try to never
let my INBOX get above 20MB either, and it has an average size of only
around 10MB or less)

(it helps to not even accept messages over 1MB in size, and sometimes
when a new worm runs rampant I cut my max message size to 150K or less
until I can work out a regex to reject the worm outright :-)

As you may know many Emacs buffer size limitations are due to the fact
that lisp pointers to objects have various bits shaved off the normal
pointer size for use as flags to the interpreter, etc., so the maximum
span of a single lisp object is a great deal less than the maximum
address space of a process (and IIRC a buffer has to fit in a lisp
object).  It's been quite a while since I looked deep enough into Emacs
guts to know what the actual lisp pointer width is on any given CPU.

However I'm now primarily using my alpha for things like Emacs VM and
with it I may well be able to open larger mailboxes (esp. since it has
1.5GB of RAM too).  I don't know though -- maybe the lisp pointer size
is no larger on the alpha?  I haven't bothered to check yet.  :-)

I don't know what the differences are between xemacs and GNU emacs
internals -- perhaps xemacs makes do with a bit or two fewer flags for
lisp object pointers.

-- 
						Greg A. Woods

+1 416 218-0098                  VE3TCP            RoboHack <woods@robohack.ca>
Planix, Inc. <woods@planix.com>          Secrets of the Weird <woods@weird.com>