On Tue, May 31, 2005 at 03:08:09PM +0200, Pierre Habouzit wrote: > > In any case, given the number of prospective ports waiting in the > > wings, 11 is probably a roughly correct estimate even if we *do* drop > > some architectures. (And since non-release ports are likely to stay > > in unstable, and adding a release port adds three w-b databases where > > dropping one only removes two w-b databases, it takes 1 1/2 dropped > > archs to balance one added arch...) > in all those debates about the lack of buildds, sth seems odd to me. > I've always thought that moore's law was quite accurate... and my > understanding (I may be wrong, it's only how I understand the whole > thing) is that debian growth is quite linear, compared to the cpu > speed, disk size, BW ... growth, the time passing, we should have less > and less limitations. Moore's Law governs the rate at which the speed of hardware (at a given price-point) doubles. It says nothing about the speed at which current software will *run* on current machines; and it certainly has nothing to say about the speed at which such software will run on machines that are no longer on the Moore's Law curve due to a lack of new hardware being designed/manufactured for that architecture. IOW, it doesn't (directly) give meaningful predictions about the rate at which a given piece of hardware becomes obsolete. It also has no capacity to predict an organization's *ability* to replace hardware. > I'm aware that more's law is not appliable on some archs (like arm I > believe) but the question is, well, who uses openoffice.org or kde on > an arm (only to cite those) ? This mitigates the linear growth of the archive itself (assuming we did subset the archive for slower archs), but it doesn't mitigate the growth of software complexity that causes subsequent revisions of the same software to run slower on the same hardware over time -- which, if it's true of nothing else, is at least true of compilers. > In fact, the point is, I don't understand how hardware limitations are > really possible. I understand fully that many ports needs *a lot* of > work for porters/security teams (it's a pity human brains does not > follow moore's law) ... but I find hard to believe that hardware is the > reason why we cannot manage many arches. I don't remember anyone ever saying this was a hardware problem. -- Steve Langasek postmodern programmer
Attachment:
signature.asc
Description: Digital signature