Peter Tribble wrote:
> On Mon, Sep 15, 2008 at 5:51 PM, Stephen Hahn <[EMAIL PROTECTED]> wrote:
>> * Moinak Ghosh <[EMAIL PROTECTED]> [2008-09-14 19:31]:
>>
>>>    2. One fundamental design approach in IPS is to use an intelligent
>>> package and metadata server. This makes IPS unsuitable for community
>>> distro mirrors. Community distros need to use public mirror services like
>>> say Ibiblio and it will be very rare, if at all for mirrors to run a custom
>>> server on their machines just to mirror a particular distro's packages.
>>  Actually, it's always been the plan to have the retrieval side be
>>  simple.  We had to take a detour when we determined that Python's base
>>  HTTP implementation didn't support HTTP/1.1, and thus couldn't
>>  pipeline.  (Trivia:  the logo actually hints at this relationship--the
>>  retrieval server is much smaller and simpler than the publication
>>  server...)
> 
> The correct approach would be to eliminate the repository/depot and
> just make packages be files that are put someplace. Whether that be a
> web server, ftp server, nfs server, CD, usb stick, or local disk is
> immaterial.

Packages are just essentially files that are put somewhere right now. 
It's just that the all the individual files of the package are put 
somewhere instead of a single file.  Once package dependencies and 
grouping has been redone properly, users will benefit from reduced 
bandwidth usage and I/O.

It is certainly our intent to eventually enable these sorts of 
installation sources.  However, our focus is currently on the network 
aspects as far as I know.

> Doing so would allow and encourage the whole world to act as a distribution
> conduit.
> 
> The current mechanism is a huge barrier to entry. Not only does it make it
> much harder for other sites to distribute your software, it makes it 
> enormously
> hard for end developers who wish to distribute software for OpenSolaris. If,
> and this will be the common case, they aren't able to run their own repository
> then they can't create or distribute OpenSolaris packages at all, which harms
> the whole ecosystem.
> 
> This is why the on-disk package file format is the crucial piece of the 
> puzzle,
> because that should be the form in which the software moves around.

I don't know why you insist that can be the only form that software 
moves around in -- it's certainly a valid form, but not the only form in 
my view.

Nevertheless, we already have a few bug points related to this:

2660 ability to install packages from "disk" (local sources)
2691 ability to publish packages to local disk repository

> (Of course, once you realize that the crux of the problem is actually making
> the software packages themselves widely available on the network, you
> realize that what's really required is not to completely reinvent the 
> packaging
> wheel, but to integrate something like pkg-get properly into the existing
> system. Not only would it guarantee compatibility with and enable reuse of
> all the existing investment by vendors and users in the existing packaging
> system, it would also allow all the benefits to be ported back to the 
> production
> releases of Solaris that Sun are going to have to support for a few years 
> yet.)

Something like pkg-get is insufficient to address needs without serious 
rework (zones, etc.).  I don't want to get into a rehash of the many 
discussions we've already had on this subject over a year or so now.

>>> Why is IPS re-inventing mirroring ?
>>  I don't believe we are.
> 
> There's a mature, widely used, and well-proven infrastructure for mirroring
> software distribution around the world. You aren't using any of it, so you're
> wither going to invent your own mechanism or not mirror at all.

If you look at our existing bug points, you'll see that we are very much 
moving towards a model where mirroring can be performed using an 
"out-of-the-box" web server (as long as it supports HTTP 1.1).

To quote j:
"The download format is going to be changed as part of other transport 
work.  Our intent is to move away from tarfiles, and instead have 
clients perform pipelined HTTP GETs."

Cheers,
-- 
Shawn Walker
_______________________________________________
pkg-discuss mailing list
[email protected]
http://mail.opensolaris.org/mailman/listinfo/pkg-discuss

Reply via email to