[r6rs-discuss] Arctic Repository 1.2 and Portable, Extensible Sockets

[Aaron W. Hsu]

On Mon, 28 Dec 2009 11:59:03 -0500, Grant Rettke <grettke at acm.org> wrote:

> On Mon, Dec 28, 2009 at 2:51 AM, Aaron W. Hsu <arcfide at sacrideo.us>  
> wrote:
>> It's also a semi-testament to the R6RS, in that
>> many of the libraries that I thought were kind of useless in the R6RS
>> turned out to be quite nice to use in practice.
>
> Please elaborate.

Basically, the record type system seemed to be overkill and the bytevector  
library didn't seem to be all that useful. The I/O system seemed to be  
really clumsy, and I didn't think I would have much use for many of the  
functions in R6RS. However, I noticed that I used almost all the libraries  
in the creation of my sockets library, including bitwise, bytevectors, and  
of course the full imperative ports system.

After having used all these functions, I have come to the conclusion that  
the bytevectors are quite useful, the I/O system, while serviceable, is  
still awkward and overly complex. Overall, it seems like most of the  
libraries in R6RS were justified, and while some of the interfaces,  
namely, input and output, are clusmy, they seem to work pretty well. There  
are improvements to be made, for sure, but the R6RS did enable to me to  
write a more portable library.

The record system is the only library on which I have mixed feelings.  
After having used it, I like it, and I take full advantage of some of the  
hairier things like protocols. On the other hand, some of these things can  
be done with a more simple interface. The result, however, isn't quite so  
compact, and it's nice to have it in a record form. Were the editors  
justified in creating such a complex record system? I'm not sure they  
were, but now that we have it, I'm not sure that I am sorry about it.

	Aaron W. Hsu

-- 
A professor is one who talks in someone else's sleep.