bkline at rksystems.com
Thu Aug 3 09:00:00 EDT 2000
On Thu, 3 Aug 2000, Jon Strande wrote:
> Okay, I may be off on a few things here, but please tell me
> if I am...
> 1.) why use a jar file on a server side application? Isn't there a
> performance hit (even a slight one) from accessing class files from a jar?
> (I do understand the ease of use issue.)
Possibly but not necessarily. There's a tradeoff between extra system
calls and I/O on one hand and additional processor cycles on the other.
The extra system calls and I/O kick in when the JVM has to get to
multiple class files and read more blocks (because the data is
uncompressed and scattered). The extra processing kicks in to do the
decompression. Usually the former costs more in degraded performance
than the latter, but you'd have to make that determination on your own
hardware to know for certain which would work better. If you had a
*really* slow (set of) processor(s) and a *really* fast I/O subsystem,
it could concievably work to your advantage to forego the jar and keep
all the class files separate. Not as safe, though, because it's easier
to have a stray out-of-date class file than an out-of-date jar.
> 2.) when the snapshot zip files are built, why aren't they packed with
> correct directory paths so that I can unzip it to my root com directory
> and just run: javac com\internetcds\jdbc\tds\*.java. When I download a
> zip, all I do is extract the source, copy the files to the correct directories and
> compile. It works perfectly, PLUS, I can skip all that makefile stuff.
Another tradeoff between convenience of getting to the source code for
editing and other tasks (diffs, for example) without navigating lenthy
paths and convenience of not having to tell the compiler where to put
the class files. Since the makefile does the latter work for you, it's
not unreasonable to opt for the former convenience.
mailto:bkline at rksystems.com
More information about the FreeTDS