Changes between Initial Version and Version 1 of Ticket #1170, comment 82


Ignore:
Timestamp:
2010-08-23T18:44:04Z (14 years ago)
Author:
warner
Comment:

Legend:

Unmodified
Added
Removed
Modified

  • Ticket #1170, comment 82

    initial v1  
    33Without more info from the 1.7.1 downloads (data which would be in the download-status, but for the old-downloader it isn't displayed until after the whole download is complete), we can't guess what servers were used for those runs. Zooko, how consistent do you think the speed-difference results would be if you used a 100MB file, instead of using the first 100MB of a multi-GB file? That might let us use Terrell's script and also collect download-status from the 1.7.1 runs.
    44
    5 It'd be awfully convenient if the speed difference that Zooko observed could be attributable to server selection, and if the combo patch made that selection work well enough to ship 1.8.0. A 1.8.1-era improvement could be to try out new servers over the course of the download, so that we'd land in the three-good-servers (sp26/nszi/4rk5) mode more often than the three-good-one-slow-servers (sp26/nszi/fp3x) mode.
     5It'd be awfully convenient if the speed difference that Zooko observed could be attributable to server selection, and if the combo patch made that selection work well enough to ship 1.8.0. A 1.8.1-era improvement could be to try out new servers over the course of the download, so that we'd land in the three-good-servers (sp26/nszi/4rk5) mode more often than the two-good-one-slow-servers (sp26/nszi/fp3x) mode.