Re: [dispatch] SIP-CLF: text versus binary

Eric,

I have pondered the questions you raise for many years, and I still 
don't know. I *think* it is mostly a matter of easing the learning curve 
to get started on something. Once you have worked on it for awhile, you 
have a set of tools, and whether the underlying representation is text 
or binary is the least of concerns. But for those first few weeks before 
you have figured out what tools you need and obtained or built them, a 
text rep is very helpful.

An interesting case study is print formats. A *long* time ago Xerox had 
a binary print representation, called InterPress. It was proprietary and 
not very approachable. Some people left Xerox, formed Adobe, and 
produced a text print format (PostScript), which for some reason was 
vastly more popular. Does it matter that postscript is text rather than 
binary? Probably not in practice, but somehow it did matter. (Or course 
now ps has been pretty much subsumed into pdf, which I think is a binary 
format.)

So, for reasons I still don't fully fathom, I agree with you that its 
probably important to have a text rep as the primary one, even if it is 
suboptimal, and if there is also a binary format people probably won't 
implement it.

	Thanks,
	Paul

Eric Burger wrote:
> I would humbly offer that the number of iterations it took for Theo 
> (super awesome coder), Adam (truly awesome coder), and Vijay (above 
> average coder) to get a half decent binary implementation of a CLF 
> writer and reader highlights the argument for a text-based LOG file.  
> Imagine what an average coder would do.
> 
> Binary file formats are not easy. SIP-CLF is trivial and it is not easy.
> 
> 
> Questions for people to consider:
> 
> Why isn't SIP in a binary format? It would be much easier to generate 
> and parse the message, and the message would be smaller.
> 
> Why is H.248 a text protocol in the wild, even though the binary format 
> had tons of intellectual effort poured into it?
> 
> Why are SNMP MIBs a black art, years after their introduction?
> 
> 
> ------------------------------------------------------------------------
> 
> _______________________________________________
> dispatch mailing list
> dispatch@ietf.org
> https://www.ietf.org/mailman/listinfo/dispatch