1 |
On Thu, May 27, 2004 at 08:09:03PM +0200 or thereabouts, Pieter Van den Abeele wrote: |
2 |
> The problem with jigdo is that it requires a network connection to |
3 |
> download the different parts that make up the iso, so it only fixes the |
4 |
> problem for people having for instance a windows box with a cdwriter at |
5 |
> work they can use to build their jigdo'ed image. (We also need a |
6 |
> windows maintainer of the jigdo code) |
7 |
|
8 |
Is it really reasonable to try and support people who can't get access to |
9 |
some sort of internet connection to download a package/iso/whatever? |
10 |
Personally, I think that's being a bit extreme. We *do not* have to |
11 |
support every possible type of installation method and I think this one |
12 |
falls into the "outside our scope" bucket, personally. |
13 |
|
14 |
> A server side solution would imho be better: have a user declare |
15 |
> requirements, submit them to a build cluster, which sends the user the |
16 |
> iso(s) by regular email or makes it availabe for download. |
17 |
|
18 |
build cluster as in build binaries? That's going to require a *lot* of |
19 |
horsepower. Far more than we have available to us now. As proof, ask |
20 |
beejay how long it takes to build all the GRP sets for the x86 subarches. |
21 |
Now multiply that by a few hundred. |
22 |
|
23 |
A more reasonable solution would be to create custom CDs of sources. I |
24 |
still think that's quite an endeavor, but it's imminently more feasible |
25 |
(imo) than trying to build the actual binaries. |
26 |
|
27 |
Again, however, I don't think trying to acommodate totally networkless |
28 |
installs is or should be within our purview. |
29 |
|
30 |
--kurt |