[chord] Re: patch comments

[Michael Walfish]

| 
| I've committed a number of modified patches based on your suggestions.
| Upon closer examination, I have the following comments...
| 
| > diff -ru chord-src-head/devel/test_types.C chord-src-noauth/devel/test_types.C
| > --- chord-src-head/devel/test_types.C	2003-07-24 01:46:21.000000000 -0400
| > +++ chord-src-noauth/devel/test_types.C	2003-07-25 04:34:56.000000000 -0400
| 
| I'm going to hold off on this one for a little while actually.
| Perhaps we can make this a more useful test program and remove
| the need for that magic2ctype stuff by streamlining the dhash_ctype
| enum. (It seems like DHASH_DNSSEC can be punted and removed by
| DHASH_NOAUTH.)

Why shouldn't the test program immediately start doing a fundamental
check, namely seeing how the dhash responds when you query for a block
that doesn't exist? I spent a while tracking down such a bug, and if there
had been a test for it, I'm not sure the bug would have existed in the
first place.

I agree the magic2ctype is hoopty, but, again, I wanted to minimize
collateral changes -- I didn't put the magic numbers in, and I wanted to
at least isolate the translation function.

The fact that the test program could be better and more useful than it
currently is (a point with which I completely agree, by the way) -- I
don't see how that's an argument for, today, rejecting a definite
improvement to it.  When the time comes to improve the program and make it
much better -- then you can rollback these changes and start over or build
on them.