Hi Alvise, On Wed, Nov 24, 2004 at 12:26:47PM +0100, Alvise Dorigo wrote: > Peter Elmer wrote: > >On Wed, Nov 24, 2004 at 12:16:28PM +0100, Alvise Dorigo wrote: > > > > > >>ok, Pete. Honestly Windows is a world to test yet... > >> > > > > No problem. At least there is something to play with. (I added Tony as > >I know he was looking at xrootd access for JAS.) > > > > > > What's JAS ? Is Tony interested in the java client ? this is encouraging... For JAS see: http://jas.freehep.org/ Regarding ROOT I/O and JAS, see his talk at CHEP: http://indico.cern.ch/contributionDisplay.py?contribId=400&sessionId=6&confId=0 IIRC, he had started to work on a Java client, but was aware (since I mentioned it at CHEP) that you were working on this... Pete > >>is >~140 kB a large attach ? ok I wont do this kind attachment in future. > >> > >> > > > > Large is all relative, but in this case the limit was set low enough to > >discourage non-trivial attachments altogether. You know how much I like > >them.... ;-) > > > > Pete > > > > > > > > > >>Peter Elmer wrote: > >> > >> > >> > >>>Hi Alvise, > >>> > >>>On Wed, Nov 24, 2004 at 11:14:07AM +0100, Alvise Dorigo wrote: > >>> > >>> > >>> > >>> > >>>>I recently fixed a number of bugs. Please use the new jar in attach > >>>>instead of that one I sent to you some day ago. Now in the source and > >>>>destination argument of XrdFileCopy you can omit the TCP_PORT number > >>>>(it will try to get it from /etc/services under linux and from > >>>>c:\\winnt\\system32\\drivers\\etc\\services under WinNT/XP, otherwise > >>>>default=1094); in addition a 3rd parameter cab be passed to specify the > >>>>size of copy chunk (larger chunk means more copy speed, default chunk > >>>>size is 256 kB) so: > >>>> > >>>>java- jar XrdCopyFile.jar xroot://<USER>@<HOST>/SRC_PATHFILE > >>>>xroot://<USER>@<HOST>/TARGET_PATHFILE [CHUNK_SIZE_IN_KB] > >>>> > >>>>Note 1: The ASYNC communication has been deeply tested with 10 > >>>>concurrent clients repeatedly reading from the same file and from the > >>>>same server (this to put all clients on the same physical channel). No > >>>>server-crash during reading has been triggered yet, but it will be done > >>>>asap. The server-crash tolerance has been fully tested with 1 and more > >>>>client in SYNC mode. > >>>> > >>>>Note 2: Some data corruption takes place when "oksofar" is handled, > >>>>that means that copying files using chunks > 2MB can create data > >>>>corruption. I'll investigate. > >>>> > >>>>Note 3: The Java client is supposed to run under WindowNT/XP too > >>>>(please remember JDK1.5) but very few tests has been done: even if Java > >>>>is fully multiplatform everybody knows that small differences are > >>>>between platforms, versions and vedors... I developed and tested under > >>>>Fedora Core 3 Linux (kernel 2.6.9-1.667) Sun JDK 1.5. > >>>> > >>>> > >>>> > >>>> > >>>Ok, thanks: I've updated the jar file linked from the xrootd page. Very > >>>nice that we have a windows client now, too.... > >>> > >>>BTW, you can't send mails with large attachments to the mailing list. It > >>>will bounce them. (And in general sending large attachments to mailing > >>>list > >>>is bad practice.) You can probably just stick any future update files > >>>someplace at SLAC and send me the location so I can put it in the xrootd > >>>web area... > >>> > >>> Pete > >>> ------------------------------------------------------------------------- Peter Elmer E-mail: [log in to unmask] Phone: +41 (22) 767-4644 Address: CERN Division PPE, Bat. 32 2C-14, CH-1211 Geneva 23, Switzerland -------------------------------------------------------------------------