On Tue, 27 Sep 2005, [ISO-8859-1] Far=E9 wrote:
> Ahem, sorry for re-sending this patch, but I got no feedback on my
> previous post. Is there anything that prevents my patch to
> sb-bsd-sockets from being included in a next sbcl 0.9.5.x ?
> It adds support for sendto, and fixes performance in recvfrom, too.
> Small test program included.
This part seems good to my socket-ignorant eye, with the exception of the
much better if it is a (simple-array (unsigned-byte 8) (*)), because that's
what's used internally and for any other type there will be lossy and/or
1) Don't expose internals in documentation. Say it is eg. fast instead.
2) The lossy/buggy comment is worrying. What sort of problems are there?
Specific issues and the way they are handled should be mentioned.
> It also includes some basic support for a rather generic type
> "iobuffer". It would be interesting to move this support to its own
> file. This support could then be built into it some better logic for
> character<->octet translation using external-format (I currently
> assume latin1), and made to be used in sb-unix and other similar
> places (I could be interested in doing that).
This part I've no idea about, given that I don't understand the use-cases.
The documentation is also a bit vague (not that I've spend much time=20
trying to make sense out of it, but a glance over left my uncertain as to=
Compounded by the fact that adding an entirely new interface is a=20
higher-cost operation (vrt. future maintenance) and the missing=20
external-format support -- I'd suggest you resubmit the first parts=20
separately for a faster merge. The iobuffer stuff can then be dealt=20
-- Nikodemus Schemer: "Buddha is small, clean, and serious."
Lispnik: "Buddha is big, has hairy armpits, and laughs."