Re: Magic numbers

Rich Salz <[email protected]> Sun, 21 Jul 2002 20:51:06 -0400
Newsgroups gmane.comp.windows.devel.dime
Message-ID <[email protected]>
> The benefit to a network protocol comes when DIME is replacing an existing
> protocol (and personally, I like the prospect and I have few candidate
> protocols in mind). In such cases, when you do not have a 'meta' protocol
> identifier, having a consistent way to recognize (or otherwise rule out)
> DIME could make it easy to switch wire protocols incrementally.

Okay, that's a valid use.  There are better -- nay, more correct --
solutions, but it's a valid use.  It's just not good enough, in my
opinion, to justify extra bytes on every DIME message just because someone
may have to transition a protocol on the same port and has no other way to
indicate it.

> Another possible benefit at the network protocol level is that it makes it
> easier to identify DIME stream in network protocol analyzers (such as
> Ethereal). I would not put a magic number just for that but it is a
> worthwhile benefit.

Funny, I'd consider that more worthwhile. :)

Only marginally so, however.  Is it common practice to use ethereal, etc.,
to just snoop on traffic without knowing what you're looking at?
        /r$