amd64 Fedora Core
Mike A. Harris
mharris at redhat.com
Tue Feb 24 17:50:20 UTC 2004
On Tue, 24 Feb 2004, Gene C. wrote:
>Justin has raised an interesting question on his weblog
>concerning Fedora core and the amd64-- should developers waste
>time on FC1 x86_64 or put their efforts into FC2 which will be
>available "real soon now"?
I can see advantages and disadvantages to both decisions. I'll
comment below.
>I can see arguments for both situations but I wonder what other
>amd64 users think.
I would assume that most AMD64 users want something useable ASAP,
so that they can use their favourite OS with their new AMD64
systems. The fastest way to get that is to release Fedora Core 1
for AMD64.
>1. With limited developer resources and FC2 test2 about two
>weeks away (it will have a x86_64 snapshot with needed i386
>packages), the better option is to put the resources into
>getting FC2 "right". With the test1 snapshot as a base and all
>of the package updates in development, the x86_64 versions
>should be in pretty good shape. Furthermore, the Red Hat build
>process is set up to handle x86_64 builds for development (and
>FC2 Test2) whereas x86_64 updated packages are few and far
>between. When the test2 snapshot rolls out, x86_64 updates will
>be available as a matter of course (or at least that is the
>plan).
All RPM packages are built on all 7 architectures, both in Fedora
Core 1, and Fedora Core 2, so from that angle, it doesn't make
any more work than we already have, to produce AMD64 packages.
There is some overhead involved with producing updates, but
_only_ if they are AMD64 specific updates. If they were updates
for x86 anyway, releasing for both x86 and AMD64 shouldn't really
pose additional effort.
For the most part, any additional effort required for AMD64,
would most likely have to be done for RHEL3, and Fedora Core 2
development also.
The only problem I see from the "work" angle, is that when Fedora
Core 1 goes end of life, it should go end of life on all
architectures simultaneously in my opinion, however if AMD64
release trails it, then either the AMD64 release isn't supported
for the same length of time from it's initial release, or else
additional effort would have to be spent supporting the AMD64
release for n months from release date. However if we supported
AMD64 for the same length of time as we planned to support x86,
then we might as well support both of them until the same EOL
date, as everything builds on both arches anyway.
Again, this is just some of my thoughts, bouncing in the air.
>2. At the current time, FC1 is more stable (especially with all
>of the updates applied) than the current state of FC2 (as could
>be expected since FC2 is still in development and testing). In
>addition, it is not at all difficult to rebuild the ix86
>packages for the x86_64 (I have done this for the current set of
>updates). FC1 test1 + updates is a fairly stable platform and
>should be pretty close to what would be available if FC1 x86_64
>final was released.
Agreed. And even after FC2 is finally released, it might be a
while before a 2.6.x based distribution could be considered as
stable as a final 2.4.x based distribution. In that case, I
would prefer myself to be running FC1 on an AMD64 box.
>So, work on getting FC2 with varying stability (there will be
>problems) or put time in to roll out a FC1 x86_64 final? At the
>present time, I lean toward option 1. I do have a working FC1
>test1 plus all updates so the release of the FC2 x86_64 final
>will be of little value to me except to do testing for the
>community.
I think you got your FC1's and FC2's mixed up in that last
sentence. Could you clarify?
IMHO, we should release FC1 for AMD64 because enough people have
spent effort on it that it would be wasted if it wasn't released.
It just wont likely have the same lifespan as the x86 release
did, because it has been released much later than the x86
release.
What is currently blocking FC1/AMD64 from being released anyway?
--
Mike A. Harris ftp://people.redhat.com/mharris
OS Systems Engineer - XFree86 maintainer - Red Hat
More information about the fedora-test-list
mailing list