Re: [PROPOSAL] New Virtual Packages and way to handle Classpath
Hallo Ean,
* Ean Schuessler wrote:
>I'm afraid that I agree with Matt. As much as I like the idea of
>abstract dependencies it seems that it will become much more complex and
>much harder than just depending on a runtime that is known to work. The
>core issue being that there are far fewer VMs than there are possible
>base class library dependencies. Easier just to say "j2sdk1.4 | kaffe |
>orp" if that is what is known to work.
And how is that differently from my aproach? j2sdk1.4 is just a word
for 'Sun compatible J2sdk of version 1.4'. mpkg-j2sdk will make a
package with that name from the sun -bin download (BD gets
j2sdk1.4-bd). So you are basicly hiding working VMs form the packageing
system. How nice, if a user has to download 30MB, just to satisfy the
dependencies and then setting JAVA_HOME to the newer -bin downlaod.
And the thing is, it will work. JAVA_HOME is not advertised anywhere,
but almost every startscript uses it. man ant, man eclipse, man
tomcat4 (oups: "No manual entry for tomcat4", which tomcat4:
/usr/bin/tomcat4). All use it, but no word in the man page (eclipse is
my package, so consider this as 'will be fixe din the next
upload'...).
Jan
--
Jan Schulz jasc@gmx.net
"Wer nicht fragt, bleibt dumm."
Reply to: