Re: [Lpr-discuss] socket.com

[James A. <ja...@va...>]

Ben Woodard <be...@va...> writes:

> James, you have that thing socket.com that you have been working on in
> your spare time. And I know you have that warning suggesting that
> people not use it for the time being. How far away from a useful state
> is it really?

 Well it was getting to be complete[1] and then I decided that it was
just sick to have a string library inside it (it wasn't that complete
at 0.3.0 ... but I kept thinking of more string like operations that
would be nice) so that's got broken out into a separate library, which
has created a necessary amount of work.
 One of the big reasons it's been going slower than I'd want is that I
didn't really have a project I could put it into (so I got distracted
by other things which actually have users, for instance :).

> gnulpr 2.0 is turning out to have some rather sophisticated network
> programming and it is likely to have a lot more of it by the time that
> it is done. It occured to me that it might be mutually beneficial for
> the production version vs. the prototype version of gnulpr 2.0 to use
> socket.com. That way all the nastyness of the socket programming could
> be encapsulated into the library and socket.com could get some good
> exercising.

 Sounds great to me :).

> What do you think? Is this a reasonable idea or do you think it would
> be better to have gnulpr 2.0 do all its own network code?

 You think I'd turn down a few million users and free debugging ... I
think not :).

[1] The only major thing that wasn't there that comes to mind is
netstrings[2] support ... but this can be implemented much easier with
the separate string library approach.

[2] http://cr.yp.to/proto/qmtp.txt is the only reference I could
quickly find (I'm sure djb wrote an article on it though).


-- 
James Antill -- <ja...@an...>
Well, it's the open source concept, but one notch better, because the
source wouldn't be open -- Derek Burney CEO of Corel
   And this guy gets _paid_!?! -- me.