[news.misc] 3.0 news -- yet more anti-confusion info

eric@snark.UUCP (Eric S. Raymond) (03/10/88)

Responses to my posting attempting to clear up the confusion about
the upcoming netnews release brought up some more points that need to
be clarified.

	1) 3.0's active file format and spool directory organization are
	   identical to those of 2.11. The sys and history file formats
	   have changed.

	2) All of 2.11's configuration options are supported either
	   identically or with close equivalents. Configuration is
	   much easier in the 3.0 version, there's a monster config
	   script that knows about all the options, asks you questions
	   and does consistency checks.

	3) 3.0 (in its earlier incarnations as Teenage Mutant Ninja Netnews)
	   has been running in production on my generic 68010 box for over
	   a year. During that time various versions of it have been tested
	   on a Sun, a Pyramid, a VAX, and a 386. The code is *very* portable,
	   all NULL-dereferences and int-equals-(char *) assumptions have been
	   long since lasered out (it lints clean!).

And then there's the *big* issue: long-term reliability.

The 2.10.X/2.11 code was getting grubby and over-hacked. Whatever design
discipline had originally been present in the B news line got lost under
multiple layers of kluging and patching as the code grew by accretion. This
devolution wasn't really anyone's fault; large systems extended by lots of
different people over time tend to get that way. But the result was hard to
modify without damage -- hence the problems that have led to fourteen official
patches and a lot of gripes since 2.11's release.

The 3.0 rewrite has hosed out all the crap and reorganized everything around a
clean architectural skeleton and an appropriate set of abstract data types. I
hope and believe that the result is considerably more robust than 2.10.X was
and 2.11 has been. At the very least, there are no longer a multitude of
strange dark cobwebby corners for bugs to lurk in; everything in the new
code is commented, all the interfaces are documented, and a debugging
exerciser is provided.

There are a couple of minor known problems in the code relating to various
storage-use and execution-time optimizations. These are documented in a BUGS
file, and I hope to have most (if not all) of them stomped by the time the
NNTP changes come back from Erik Fair.




-- 
      Eric S. Raymond                  (the mad mastermind of TMN-Netnews)
      UUCP:  {{seismo,ihnp4,rutgers}!cbmvax,sdcrdcf!burdvax,vu-vlsi}!snark!eric
      Post:  22 South Warren Avenue, Malvern, PA 19355    Phone: (215)-296-5718