From: <jm...@st...> - 2000-05-20 21:10:57
|
Sam Steingold <sd...@gn...> writes: > "John M. Adams" wrote: > > > > Sam Steingold <sd...@gn...> writes: > > > > > I would like to switch to CLOS in clocc/src/port/net.lisp. > > > This means that both SOCKET and SOCKET-SERVER will be CLOS classes > > > and all functions operating on them will be generic. > > > > How would a closified net.lisp be better? > > Come on, this is the old "OO" vs. "traditional". > > OO has the usual advantages of extensibility and maintainability > and disadvantages of size and performance (should be irrelevant here: > performance of the network is always the bottleneck, not the performance > of the code). Oh, so this is more of a let's clossify for abstract reasons as opposed to we need to support multiple protocols and having classes would make that easier kind of a change? I don't necessarily have a problem with that. Clossifying net.lisp will not automatically make it better in any sense. That's why I asked the question. I thought maybe you had something concrete in mind. In any case, since you're the author, if clossifying it makes it better for you, well, I'd say that's a good thing. The systems I work on these days rely on clos and the mop and I like it that way. In other words I'm clos-friendly. Before that I spent several years working on large distributed systems (Tivoli orbs). Among other things, I worked on the nitty-gritty comm code including various interface layers that gave us a portable, thread aware, comm api over sockets, tli, et. al. on Windows and Unix. This is my perspective. What sorts of applications are currently using net.lisp? -- John M. Adams |