On Sun, Feb 23, 2003 at 03:44:53PM -0500, Jay Oliveri wrote: > On Sunday 23 February 2003 02:37 pm, Ian Clarke wrote: > > What is the point in forcing every client author to reimplement FEC > > when the FEC code is already present within Fred? > > It is kludgy to say the least if you're inserting data into Fred from > another machine. The data sent back and forth over the wire (even through > localhost) is simply unecessary and wasteful. > > Even though most inserts are sent to localhost, inserts theoretically can > occur over a group of nodes all simultaneously (I think fishtools does > this). > > Also, using Fred to do FEC encoding means Fred is now extremely busy, > meaning I cannot FEC encode another file at the same time with the same > node. By moving the encoding offline, in the long run inserts will be > faster.
You can implent FEC yourself if you want. I've done it before, and used
to do things this way. THe gain isn't as large as you think.
- fish
pgp00000.pgp
Description: PGP signature
