On Sat, Oct 05, 2002 at 11:57:23AM -0400, Gianni Johansson wrote:
> On Wednesday 02 October 2002 19:23, ian wrote:
> 
> > > Shouldn't the datastore be used for this kind of thing?
> >
> There are four issues:
Now that the native datastore is the default, why can't we default the
temporary bucket dir, for both FEC and fproxy, to store/temp/, and
create this dir?
> 
> 1) The BucketFactory interface doesn't allow you to create a bucket of 
> unknown size.  As a result, you have to either massively overallocate or rely 
> on unspecified behavior of the underlying implementation.  The FEC 
> inserting/downloading code in fproxy does latter.
> 
> i.e. bf.makeBucket(-1) // Works only because the BucketFactory implementation 
> is making FileBuckets.
> 
> 2) The FSBucketFactory.FSBucket.size returns the requested bucket size when 
> the bucket was created, *not* the number of bytes that have been written into 
> the bucket.  This is a violation of the Bucket interface semantics and is 
> just wrong, but I guess it's hard to fix (maybe it's easier with the new 
> native datastore implementation?) 
> 
> I argued with Tavin over this for a long time last year and finally just gave 
> up.  You have to know how to pick your battles :-(
> 
> The effect is that even if you massively overallocate, hoping that you will 
> have enough space for the data for the key you are requesting, there's no 
> what to know how much data you actually got.
> 
> 3) FEC plugin implementation assumes FileBucket derived Buckets.
> 
> This is a hack. My bad.  But without it striping would work orders of 
> magnitude slower.
> 
> The issue is that you need to be able to map a Bucket to a File (containing 
> unencrypted data) so that you can seek().
> 
> 4) The shear size of FEC temp files.
> Do you really want users to flush their data store every time they insert an 
> ISO?
> 
> I haven't been keeping up with the list.  Can the new datastore grow without 
> bound?  If not, the datastore size would impose an upper limit on the maximum 
> FEC splitfile size.
> 
> I guess it wouldn't be that hard to hack your way around 1 and 2,  and 4 
> (separate datastore for fec temp files. yuck). 3 is much tougher.
> 
> > Just how are these settings intended to be used?
> >
> 
> # Make sure you use a trailing slash.  I guess I should make it optional.
> 
> #Set this parameter to a directory with a lot of free space.
> FECTempDir=/home/giannij/freenet_module/fectmp/
> 
> #Set this parameter to a directory with a lot of free space
> mainport.params.servlet.1.params.tempDir=/usr2/giannij_space/fproxy_tmp/
> 
> These can point to the same directory.  Ideally we would have only one 
> parameter, but you would have to do some sort of rude hack, given the 
> different manner in which fred and servlet parameters are handled.  i.e. 
> Ideally servlets should know nothing about fred parameters so they can run in 
> an abitrary Container.  We have violated that principle in so many other 
> places, I guess one more hack woudn't hurt.
> 
> Also, I am somewhat reluctant to do anything that requires changes to 
> parameter handling after the fiasco of the week before last.
> 
> --gj  
> 
> p.s. 
> Your README.paranoid file is a good idea.
> 
> People who are really bent out of shape about unencrypted temp files should 
> consider whether their JVM is using locked memory...
> 
> > Ian.
> >
> > On Wed, Oct 02, 2002 at 04:20:34PM -0700, Reuben Balik wrote:
> > > What would be the recommended entries into the
> > > freenet.ini to fix the following errors:
> > >
> > > WARNING: FECTempDir not set in freenet.conf/ini!
> > > Oct 2, 2002 5:31:55 PM (freenet.node.Main, main):
> > > WARNING: FECTempDir not set in freenet.conf/ini!
> > > Oct 2, 2002 6:15:44 PM
> > > (freenet.client.http.FproxyServlet, QThread-10):
> > > WARNING: fproxy tempDir not set.
> > > Oct 2, 2002 6:15:44 PM
> > > (freenet.client.http.FproxyServlet, QThread-10):
> > >    Set mainport.params.servlet.1.params.tempDir in
> > > freenet.conf/ini.
> > > WARNING: fproxy tempDir not set.
> > >
> > > --Reuben Balik
> > >
> > > __________________________________________________
> > > Do you Yahoo!?
> > > New DSL Internet Access from SBC & Yahoo!
> > > http://sbc.yahoo.com
> > >
> > > _______________________________________________
> > > devl mailing list
> > > devl at freenetproject.org
> > > http://hawk.freenetproject.org/cgi-bin/mailman/listinfo/devl
> 
> ----------------------------------------
> Content-Type: application/pgp-signature; charset="us-ascii"; 
> name="Attachment: 1"
> Content-Transfer-Encoding: 7bit
> Content-Description: 
> ----------------------------------------
> 
> _______________________________________________
> devl mailing list
> devl at freenetproject.org
> http://hawk.freenetproject.org/cgi-bin/mailman/listinfo/devl
> 

-- 
Matthew Toseland
toad at amphibian.dyndns.org
amphibian at users.sourceforge.net
Freenet/Coldstore open source hacker.
Employed full time by Freenet Project Inc. from 11/9/02 to 11/11/02.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: not available
URL: 
<https://emu.freenetproject.org/pipermail/devl/attachments/20021015/6967ee00/attachment.pgp>

Reply via email to