[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: port? in R5RS

I agree with Richard the answer to Sergei ought to be that PORT? is
not part of that environment since in his own owrds "it is not

But there's a lesson or two in here, whether we choose to heed them or not.

 - One simply should not make a zillion edits and then release 
   the next day.  This is not a slight on our editor--any editor could
   make a mistake doing that much work at last minute; just the opposite,
   the schedule should allow time enough for the editor's work to be 
   reviewed so the editor isn't responsible for changes made at the last
   minute.  (I'm not assuming the port? error was made at the last minute,
   btw.  I don't know that it was.  But it easily could have been.)

 - A specific period of time during which specific anticipatable
   quality assurance tests are done (perhaps divvied up to each author
   or to authors in pairs or something--might as well give those of us who
   claim credit some specific responsibilities as well; better than us
   each choosing tasks on our own that perhaps duplicate effort in one area
   while leaving another uncovered).  Things that come to mind are: 
   correctness of index, correctness of examples, correctness of semantics,
   completeness (all named symbols that should be defined are defined), etc.
   A number of these could be automatically tested or semi-automatically 
   tested, with a little creative effort.  But even manually checking on a
   document this small isn't unbearable.

 - A beta test period during which the public could preview things and catch
   what we did not think to test might be good.  This might be especially 
   important in R6 if we're thinking (as some have suggested) of having it be
   the last.  I'd hate to see an R7 whose sole purpose in life was to fix a bug
   noticed only moments after R6 was published.

Yes, such processes do increase the degree of bureaucracy and red tape
required to put the document together.  But people don't create such
processes to slow things down--they make them to solve or avoid
specific problems that commonly come up.  If there's really a public
out there clamoring for a revised spec, then presumably that public
also wants the spec to be correct.  (Otherwise, they'll just be
clamoring again immediately for a re-revised spec...)

I don't expect anyone's going to agree with me on any of this, which
may seem like "overkill" and "too late" at this point.  But I felt a
need to say it "just for the record"...