Jim Meyering wrote: > I'm surprised you'd compare in such a pessimistic manner.
The obvious conclusion is that I am a pessimist! :-) More seriously though I think that is the proper way to compare the data. We are talking about how well compression works which means we have to take into consideration the data that is being compressed. > I look at it differently: > > compare download time: > going from gzip to lzma, I see a speed-up of 2.39 > going from bzip2 to lzma, I see a speed-up of 1.55 But when talking about times that are less than a minute is it significant? Not to me. I am sure that my bandwidth wasted to spam swamps it. > compare disk usage: > going from gzip to lzma, I can store more than twice as much (2.39x) data > going from bzip2 to lzma, I can store 55% more data A 100G disk could hold 10,000 copies of 10M projects. Compressed source code projects are not usually what fills up the typical disk. Being able to store 2.39x more projects means 24,000 copies. A 100G disk is now small. Being able to buy a 500G disk for $100 means being able to store 50,000 copies. Combined is 120,000 copies. "Enough is equal to a feast." I won't be disk space limited due to the relatively small difference in compressed distribution files. Alternatively a git clone contains the complete history (what was converted to it) and rests at 53M. In terms of efficiency it can reproduce any version in the history very efficiently. Comparing lzma to gzip does not produce the huge benefit that is obtained when comparing, for example, git to cvs. That is clearly so much improved that it is worth the cost associated with it to move to it. I am not unhappy with lzma. It seems quite reasonable. I am happy to use it. I am just providing a counterpoint to balance some of the discussion. But I am not wanting to be too quarrelsome about this (just a little bit but not too much) so I will quiet down about it. :-) Bob _______________________________________________ Bug-coreutils mailing list [email protected] http://lists.gnu.org/mailman/listinfo/bug-coreutils
