Branden Robinson <[EMAIL PROTECTED]> writes: > On Fri, Dec 14, 2001 at 04:41:52PM -0800, Thomas Bushnell, BSG wrote: > > Branden Robinson <[EMAIL PROTECTED]> writes: > > > > > It's been discussed to death. Some people want to be able to include > > > megabytes upon megabytes of invariant non-technical sections in main. > > > Others don't. > > > > Has anyone suggested that? I haven't. > > You suggested a proportional limit, so there is nothing to rule out > megabytes of invariant text, if either or both the allowed proportion > or the package's raw size are large.
You've misunderstood two different statements I made and treated them as interchangeable. My preferred standard would only permit reasonably small amounts of invariant text. From your quote above, I expect you think that "megabytes upon megabytes" would not be reasonable. I agree--it would not be reasonable. Therefore under the "reasonably small amounts" test, it would fail. But, if my preferred standard, the non-legalistic one, is not likely to gather support, then I would be willing to see something like your earlier proposals, but with a replacement of the byte-count-per-package limit by a percentage limit. Under your proposals, in which you have a byte-count limit, there is nothing to prevent the inclusion of megabytes upon megabytes of invariant text in main--this is true whether one uses a per-package bytecount or a per-package proportionality test. I just think a proportionality test is more appopriate. While both tests can be worked around, it's harder to work around a proportionality test. Working around a per-package byte-count requires merely splitting things into different packages. Working around a proportionality test requires the integration of real content into Debian. This requires Real Work and not mere splitting-into-multiple-packages, so it's harder and therefore less likely to happen. Also, if someone does do it, they have to make a proportionately large contribution of real stuff to Debian, which is some sort of offsetting gain. This is a technical point, not a political one. My preference for a proportionality test is not any kind of attempt to make your proposal go away; it's a simple technical fix of a certain kind of problem that I envision, and one which, in fact, would make your proposal entirely satisfactory to me. (Assuming we set the percentage appropriately; my guess is that something like 0.1% is going to be more-or-less right, though I'm not sure.) Division into packages is a somewhat artificial thing, in other words--it's an artifact of technology and not an indication of anything "real" in the content being packaged. Accordingly, guidelines like the one your proposal is concerned with should not use "per-package" as the metric; I simply want to replace that with "per-bytes-of-content" so that it tracks something that isn't so arbitrary. Thomas
