csw@ulysses.homer.nj.att.com (Chris Warth) (03/29/88)
BUGS in NeWS version 1.1 - or -
Comments from the peanut gallery.
We *FINALLY* received our copy of NeWS version 1.1. It was really
frustrating to hear about all the neat things in 1.1 from the
seemingly hundreds of Beta-testers while we couldn't even get a taste
of the new stuff. Unfortunately version 1.1 is a let-down. Oh, they
did a good job of tracking down bad pointer references, and Gosling
made an attempt at implementing the Adobe font machinery, but they
fell far short of coming up with a quality product.
There are pervasive type errors in NeWS that make it very
difficult to port working PostScript programs. Why is a font in NeWS
of type "fonttype" while a font in PostScript is of type "dicttype"?
Adobe clearly states that a font is a dictionary.
Even worse, the contents of the "fonttype"s is wrong. On page 3
of the "NeWS 1.1 Release Notes", about halfway down the page, Sun
states, "Full Font Model Now Supported [...] definefont now works".
Well, sort of. definefont does work, but not on any of the NeWS
fonts... Y'see, in the NeWS fonts the Encoding vector (read "array")
is actually a dictionary. definefont correctly expects Encoding to be
an array and fails with a typecheck error if used on NeWS fonts. A
related problem is that all the fonts are missing a CharStrings
dictionary. Sun even has the gall to suggest one "read the Adobe
specification carefully" to justify the inoperability of charpath. I
would suggest *you* read the Adobe spec, Sun, and redo the fonts so
they they conform to it.
Version 1.1 still believes that ".1234" is a wholly different type
of object from ".12345". This is fine if the difference were
invisible to the user. They did modify the printing routines to
consider these two types as equivalent, but they forgot to tell just
about every other routine that can take a real number. One way to see
it is to try the following:
.1234 .1234 .1234 sethsbcolor
<works fine>
.12345 .1234 .1234 sethsbcolor
<fails with a typecheck error>
BTW, this error is not limited to numbers read in by the parser;
mathematical calculations can also come up with these numbers that
many operators don't recognize. You must always be careful to
truncate your results to four decimal places before using them in
NeWS. Don Hopkins must have the patience of a saint to put up with
this kind of crap and still do groundbreaking work in NeWS.
What is really galling, though, is that while substantive errors
like the above remain, Sun has time to fix things like bug #1004628,
"anthromorphism [sic] of the start up message 'I' is inappropriate."
'Nuff said.
Chris Warth
ATT Bell Laboratories
Murray Hill, NJ
ulysses!csw
--
P.S. These comments are just that, comments from a NeWS user and
application developer, not someone who actually has to fix the
above problems; I do have to live with them and find workarounds,
though. Until something better comes along, I won't stop using NeWS,
but I sure would like to see it live up to its potential.
--