On Sat, Apr 14, 2007 at 12:30:53PM +0200, Florent Daigni?re (NextGen$) wrote: > * bbackde at googlemail.com <bbackde at googlemail.com> [2007-04-14 10:32:25]: > > > I read the wiki doc about TestDDA and I have some comments/questions. > > > > 1) > > Please define "READ" and "WRITE" exactly. > > > Do you plan to do any read operation? > > > > After some time I came to the conclusion that "READ" means I want to > > GET/request a file from freenet, and the node have to write it into > > the directory. IMHO READ/WRITE with the opposite meaning is to > > abstract for a client developer *g*. But at least define READ^=GET and > > WRITE^=PUT to make it clearer. > > > > You got it right : READ= get a file from freenet
I thought READ meant *read from the disk* ? Isn't that the simplest terminology? > > > 2) > > >You can retry to enable DDA requests more than once for either the > > same or a different directory. Only last test result will be taken > > into account. > > > > Please describe in more detail. Can I TestDDA several directories at > > the beginning of my application run, because I use multiple > > directories? Or is always only 1 directory allowed for DDA? And how > > does the node remember what client tested which directory, is this > > bound to the socket connection of the client and must be repeated for > > each new connection? > > You can TestDDA as many directories as you want when your application > starts; the node handles it per socket, meaning that you have to repeat > it on each new connection... And you can't replay TestDDAResponses. > > > 3) TestDDAReply > > > > > You *SHOULD* send a TestDDAResponse message after getting it. > > > > *SHOULD* or *MUST* ? I think you must. > > Well, if you don't the node won't clean up the files, so yes I guess > it's a must. > > > And I miss some words that the client have to create a new file and > > write the WriteContent into it. How should a client behave if the file > > creation fails ? Just skip the creation and let the node recognize > > that the file is not there? > > Yes, the test will fail; the client ought to fallback to direct access > mode. > > > 4) Windows vs. *nix > > Maybe add some words that the TestDDA concept is bound to the > > directory name completely. Mapped network drives with different names > > (e.g. on linux and windows) could be DDA eligable, but due to the > > directory name used in TestDDA you cannot use DDA there. > > There is no crossplatform way of identifying them (thanks to windows not > respecting URI conventions) ... and java isn't able to handle that anyway. Absolute pathnames. They include the directory and the drive. > > On linux you don't map a network drive; you mount it. And mounted > filesystems will usable with testDDA provided the mount point path is > the same... I guess you can do the same on windows mapping the same > drive letter. > > > I will do some tests with the new code. Is it really true that 1028 > > will make this mandatory? In all other cases you gave us a specific > > amount of time instead a fix next version number. What if you need to > > release 2 version in the next few days? > > Well, the API is stable now (me and toad agree on it) so it's unlikely > to change... 1027 won't enforce testDDA but will feature it; 1028 is > planned to be a self-mandatory release so yes it will enforce it... and > no, sorry there is no ETA appart from "soon" yet. Why do you want 1028 to be self-mandatory? > > I will update the wiki to take your remarks into account :) > > NextGen$ -------------- next part -------------- A non-text attachment was scrubbed... Name: signature.asc Type: application/pgp-signature Size: 189 bytes Desc: Digital signature URL: <https://emu.freenetproject.org/pipermail/devl/attachments/20070414/2acc4c3d/attachment.pgp>
