Mercurial > hg > ventivac / changeset: todo with things that may be interesting in the future. readme now has alternate web page url and my e-mail addresses.
||Mechiel Lukkien <firstname.lastname@example.org>
||Mon, 20 Aug 2007 15:39:21 +0200
||todo with things that may be interesting in the future. readme now has alternate web page url and my e-mail addresses.
1.1--- a/README Mon Aug 20 14:58:51 2007 +0200
1.2+++ b/README Mon Aug 20 15:39:21 2007 +0200
1.3@@ -18,10 +18,24 @@
1.6 the programs that are installed by the mkfiles all have manual pages.
1.7-except for appl/lib/venti.b and man/2/venti (which are from inferno-os
1.8-svn), these files are probably MIT licenced as recommended for all plan
1.9+except for appl/lib/venti.b and man/2/venti (which are originally from
1.10+inferno-os svn but have been modified slightly), the files in this
1.11+repository are under a standard MIT-licence as recommended for all plan
1.12 9 google summer of code projects.
1.14+the mercurial repository will be used for future changes, it can be
1.19+additionally, i will keep a page about ventivac updated on my own
1.24+if these were both to fail, you should be able to reach me at either
email@example.com, or firstname.lastname@example.org.
1.27 other notes:
1.29 - appl/cmd/testrabin.b reads a file and splits it using appl/lib/rabin.b.
2.1--- /dev/null Thu Jan 01 00:00:00 1970 +0000
2.2+++ b/TODO Mon Aug 20 15:39:21 2007 +0200
2.3@@ -0,0 +1,5 @@
2.4+some more ideas related to ventivac:
2.6+- a venti write buffering proxy. it should return the score written immediately, and write it to a remote server in the background. this makes synchronous writes on a low-latency link look fast. a sync would wait until all writes have finished, so there is no risk of losing data.
2.7+- a better cache replacement algorithm for vcache could be implemented. more testing of the performance of the current clock algorithm should be done first.
2.8+- a venti-proxy that does read-ahead on the raw venti blocks. 9p-level read-ahead will help for e.g. vacfs. for vacput with writing compared to an existing venti archive, a 9p-level read-ahead proxy would not help, but this would. questionable whether it's worth the trouble though.