Re: i can't win...

>> The only missing statistic from the the comparative study, IMHO, is
>> .ZIP.
>> If all you're after is lossless compression large image files, .ZIP
>> may
>> well provide results comparable to Zonn's format.

ZIP creates files, that for the most part, are at least twice as big as .BMZ,
usually even bigger.
 
>Yeah, I dunno on that one. LHA, ARJ, and the like might do better or
>worse I suppose...

None of the "universal" compressors will compete with .BMZ, though there is a
standard I found out about after I was finishing up with .BMZ. It's called JBIG
and it uses the same core algorithm for compression. Unfortunately the people
writing it are (IMHO) very unix biased. There is source code available for it,
but it will only run on a 32 bit unix system and the protocol is very
complicated. The source says: "Not tested on any other machines but there are
sure to be problems if 'int' does not = 'long int' on your compiler."

It's being designed to do progressive downloads (like .GIF and .JPG) and so far
nothing runs in anything PC (that I found -- I didn't look that hard). Like it
or not I believe that's a mistake on their part. Because of its progressive
nature, and all the stuff dealing with on the fly decoding it's memory usage is
very high, and the source was very hard to follow. Only the core is similar to
mine beyond that the algorithm gets complicated by allowing progressive
downloads.

As far as just plain compression I'm sure BMZ will do a better job. I use a
bigger model state (I look at more surrounding pixels to determine the best
guess at what the current pixel is). And I use real multiply and divide
routines. JBIG uses a shift algorithm that is a heavily enforced patent of
IBMs. It's great for small processors that have trouble doing MUL and DIV, on a
pentium these instructions are pretty darned fast and not much is lost in speed
by using them, and a bit of compression is gained.
>
>> One caveat - if Zonn's format contains optimizations for long vertical
>> lines as well as long horizontal lines, it may well outperform .GIF
>> and
>> .ZIP when applied to schematic drawings.

Ok, I'll write up a simple document describing the algorithm, it's basically
pretty simple and was published back in 1981 as a competitor to the then being
considered FAX standard (there was no source given with the article). I've
tweaked the algorithm implementation a bit and converted it to real code.
 
>Someone should write a .BMZ plugin for Netscape and IExplorer. ;-)

I thought of that, but JBIG really is the way to go there, and I'm not sure if
the newest releases of both won't already have JBIG in them -- the JBIG people
are really lobbying for it. It could depend a what IBM wants for its patent.

(BTW: Stac recently announced they own the patent on the PKZIP algorithm, which
is pretty funny because the patent is dated something like 10 years after the
publishing of the algorithm in 1977, and after PKZIP was already using it --
PKZIP also claim a patent on their version of the algorithm. By it's own
admission the patent office has no idea of how to deal with software patents and
is currently excepting all properly word patents and letting the courts sort
them out.)

>> (This may only hold true for cases where the scan was performed with
>> the
>> schematic being very well lined-up with the scanner during the
>> scanning
>> process. For fun, take a .GIF of a schematic and rotate it by 2-3
>> degrees, and watch the size grow...)

.GIF would have a harder time with that than .BMZ would.
 
>Ahhhh, good point, but at sufficiently high resolution even an uneven
>scan will probably have vertical or horizontal lines that are made up of
>multiple-pixel widths.
>
>Of course this is kind-of amusing to me since I think the last IDE drive
>I got for my PC has like 64K sectors or something anyway. 652 bytes
>worth of schematic? 64K of file system... ;-)

Right! But on my home page 652 bytes are 652 bytes when it comes to the end of
the month billing. The funny part is those 652 byte probably do take up 64k of
their file system. Not my problem! ;^)

-Zonn

<><><><><><><><><><><><><><><><><><><><><><><><><><><><

 ------ ___ Member of A.A.C.S.:
 |---- | ( ) Association for Artistically
    / / ( () ) Challenged Signatures
   / / //\\ // (__)
  / ---/ // \\ //\\ // zonn @ zonn . com
 -------| // \\/
Received on Fri May 15 15:02:53 1998