----- Original Message ----- From: "Gianni Johansson" <[email protected]>
> On Wednesday 29 January 2003 16:21, you wrote: > > ----- Original Message ----- > > From: "Gianni Johansson" <giannijohansson at attbi.com> > > > > > I thought I had mailed this link to the list a while back, but here it is > > > again. > > > > http://cvs.sourceforge.net/cgi-bin/viewcvs.cgi/freenet/Contrib/gj/FCP-FEC.t > >x t > > > > > This whole episode with inserting metadata into data blocks should have > > > > never > > > > > happened. I tried to get the following changes to the SplitFile spec > > > made > > > > two > > > > > weeks ago. > > > > http://hawk.freenetproject.org:8080/pipermail/devl/2003-January/003373.html > > > > <snip> > > > > > 1) THIS SHOULD BE ADDED TO THE SPEC. > > > > -- 1) Specification of cipher used for all block CHKs > > -- SplitFile.BlockCipher=TwoFish > > > > I am wondering why exactly this should be added. It aids with healing given > > abilities a client does not usually have. > The FCPServer implementation definitely reads the cipher out of the > ClientPut.Cipher field. > > See: > http://cvs.sourceforge.net/cgi-bin/viewcvs.cgi/freenet/freenet/src/freenet/n ode/states/FCP/NewClientPut.java?rev=HEAD&content-type=text/vnd.viewcvs-mark up > > Unless I am missing something, it does look like you can specify the cipher. You seem to be correct. The FCP docs should be updated to show this. > > We cannot choose the block cipher > > to encrypt with. (unless things have changed) > > > > Counterproposal: > > We could make the blanket statement that all the blocks of the splitfile > > need to be the same encoding. It would have the same end result, assuming > > that atleast one block was retrievable. > Why make things complicated. What do you gain by not explictly stating which > cipher is used? We gain that the clients don't need to override the node's defaults on what the cipher is. True that we can assume that we want to use TwoFish, but not having defaults like that may make things nicer. To set BlockCipher=TwoFish, the client inserting explicitly has to set it on the insert of the control doc and the insert of all the blocks - it may make things more simple to assume that since there is a default in fred, go with that and only check things like that on rehealing. I'm not sure, but it doesn't seem to complicate things much more than any other complication we could add and doing it would allow the client more metadata while it was at it. *shrug* > --gj -Mathew _______________________________________________ devl mailing list devl at freenetproject.org http://hawk.freenetproject.org:8080/cgi-bin/mailman/listinfo/devl
