> So... that is the design we happen to be using in Apache, and is emboded in > the various APR and APRUTIL bits. It might be interesting to determine > whether a similar design would work for Samba's needs.
samba is an incredibly large monolith. it is a combination of twelve to fifteen separate entities, in only two programs, totalling 200,000 lines of code (the rest is client-side programs). to introduce some newbie programmer with lots of enthusiasm, or even an experienced programmer with lots of expertise, into these 350,000 lines, is asking a lot. hence the TNG initiative, which subdivides along sensible lines into separate programs and separate projects. elrond, for example, has adopted lsarpcd, which is only 15,000 or so lines (at a guess: he's done a lot of work on it recently). the buckets idea sounds like it is very fast, however my impression is that it is asking programmers to follow some conventions [for decoding] that an API would impose on them. the buckets idea sounds similar to what ISS proposed for a redesign of their RealSecure Engine, which would do multi-layered, multi-threaded decoding of raw ethernet traffic into first ICMP and UDP, then hand-off to handlers to do UDP to TCP conversion, then hand-off from the UDP and TCP "buckets" to protocols such as SSL, http, pop3 etc etc, then hand-off from SSL to http, pop3 etc etc: you get the idea. all very fast, all in one program, all very integrated. the TNG approach is to divide and conquer, which makes for scalability in project management. idea. is it possible to wrap the buckets idea with this network abstraction layer idea? i.e. to implement the NAP using the existing APRUTILS buckets code? luke ----- Luke Kenneth Casson Leighton <[EMAIL PROTECTED]> ----- "i want a world of dreams, run by near-sighted visionaries" "good. that's them sorted out. now, on _this_ world..."
