Daniel Carrera wrote:
John W. Kennedy wrote:
12/05/2005 10:49 PM 69,999,781 test1.xml
12/05/2005 10:53 PM 26,167,179 test1.zip
12/05/2005 11:05 PM 167,999,781 test2.xml
12/05/2005 11:09 PM 28,641,918 test2.zip
Clearly, the size of the tagname is fairly unimportant.
Bingo. The tag names are easily-removed redundancy and are compressed
away both on disk (by ZIP) and on memory (by pointers).
Cheers,
Daniel.
God, Daniel, that doesn't prove a thing! I *know* the XML is redundant,
I *know* it's zipped on the disc, and I *know* that it doesn't exist
except during the file loading and saving processes. But what you seem
to be saying is that the flat XML never exists anywhere at all, for any
length of time. Not on disc, not in memory, nowhere, for any length of
time, however brief.
In fact, I have no idea what John's point was supposed to be with that
exercise. The file I was talking about had an even more impressive
compression ratio, about 60 to 1. John, in another post and I thank him,
actually comprehended the point I've been trying to make. If the file is
zipped on the disc, then you aren't reading the flat xml from the disc;
you're reading it from an image of the inflated file in RAM.
So the only question, and John nailed it perfectly, is whether or not
the inflated file image does or has to exist whole in memory to parse
it. Hence, my question concerning whether the content.xml is parsed
serially. If parsing the XML file is a one-pass operation, then it seems
that the operations of inflating and parsing could be performed in
parallel threads dramatically reducing the memory demands. There seems
to be no concensus on that point, or maybe you all just don't know.
So if that all is true, then what we probably really have going on here
is that OOo processes the file inefficiently, and in a fashion that
makes heavy demands on memory, and that inefficiency is only exacerbated
by the file size, which *is* a direct mathematical consequence of the
verbosity of the tags.
BTW, have you looked at the subject line lately? Talk about drifting!
--
Rod
---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
