Jeff Johnson wrote:
Alan Cox wrote:
On Tue, Feb 22, 2005 at 10:40:34AM -0500, Demond James wrote:
I say let's look elsewhere before we start swinging the axe at Java stuff. We can still keep them on the list, but I say start with the bigger games and the redundant applications first.
The few really redundant applications we have don't make a tiny dent in the
space required for java. The more I think about it the more it seems to come
down to "Gnome, KDE, Java, Open Office" pick any three.
I happen to agree with 3 out of 4, makes perfect sense to me.
However, I'm obligated to point out that there is some mechanical drudgery
that might put off the day of reckoning for what packages should be in
FC4.
The space constraints are approximately 4*650Mb = 2.3Gb.
Eeep, serves me right for multiplying in public, sigh. s/2.3/2.6/
Current overage is 300Mb, so package real estate is currently (estimated) 2.6Gb.
s/2.6/2.9/
Headers are (or were) ~12-15% of package real estate.
Let's use 10% for the analysis, or 260Mb of headers in current FC4.
s/260/290/
Compressing headers would save about half of that, or ~130Mb, more if changelogs
were truncated during build.
s/130/145/
Much learned discussion (jnovy in particular iirc) points out additional savings achievable
by choosing to use bzip2 for certain large package payloads. I'll wave my hands here,
but I'm pretty sure that a big chunk of 170Mb could be saved.
s/170/155/
Yes new rpm features, but zlib ain't exactly hard coding, nor is a date comparison loop
for truncating changelogs, nor is configuring bzip2 payloads for certain package payloads.
And yes, there's a knapsack problem fitting packages onto 4 CD's in priority order
that is not addressed above at all.
Again, my personal belief is that 3 out of 4 is sounder (as in soundboard) starting point
for FC discussion.
The analysis is more important than the computational details ...
And 3 out of 4 is still my personal choice.
Sorry for the confusion.
73 de Jeff
