Brian Warner <[email protected]> writes:

> On 3/5/12 11:38 AM, Greg Troxel wrote:
>> 
>> No argument about figuring things out, but from a security viewpoint
>> it would be nice to have a tahoe server process that does not have the
>> client code even loaded. (I realize there is a notion of a WUI and a
>> node that is client and server, but I prefer to run nodes that are
>> clients and nodes that are servers and keep them totally separate.)
>
> Yup.
>
>> FWIW, on NetBSD 5.1ish i386, python 2.6, I have a process running an
>> introducer, 33K size, 13K RSS, and a storage node, 52K size, 20K RSS.
>
> 33 *K*? Surely you mean 33 megabytes (maybe 33k pages of 1kB each, or
> something?).

Sorry, I did indeed mean MB.  I had just been reading about
retrocomputing too recently and must have been thinking of the good old
days, when programs that fit in 33K could do something useful.

> I think the tahoe Introducer code path imports a lot less code the
> client/server node. Changing the structure to split client from server
> in a similar way would probably help.

That would be nice.  But, I was surprised that the server was only
52M/20M (and still surprised that I could say "only 52M").

Attachment: pgpXRnfAInTgd.pgp
Description: PGP signature

_______________________________________________
tahoe-dev mailing list
[email protected]
http://tahoe-lafs.org/cgi-bin/mailman/listinfo/tahoe-dev

Reply via email to