[Date Prev][Date Next][Thread Prev][Thread Next][Author Index][Date Index][Thread Index]

Re: [Gzigzag-commits] gzigzag/src/mediaserver SimpleMediaserver.java



On Thu, Jul 26, 2001 at 12:51:35PM +0200, B. Fallenstein wrote:
> 
> Tuomas Lukka wrote:
> > On Thu, Jul 26, 2001 at 12:38:42PM +0200, B. Fallenstein wrote:
> > >
> > > "Tuomas J. Lukka" wrote:
> > > > Update of /cvsroot/gzigzag//gzigzag/src/mediaserver
> > > > In directory usw-pr-cvs1:/tmp/cvs-serv30681/mediaserver
> > > >
> > > > Modified Files:
> > > >         SimpleMediaserver.java
> > > > Log Message:
> > > > Ah: the reason for the slowdown found. SimpleMediaServer must cache blocks.
> > >
> > > I haven't looked at the code yet, but I do not agree with what you say
> > > here: in order to keep functionality simple, SimpleMediaserver must
> > > _not_ cache blocks. That should be done in interaction with other
> > > mediaservers (e.g. MultiplexingMediaserver is there for this very reason).
> > 
> > Something below GZZ1Reader must cache blocks. Doesn't atter what class
> > it is. Currently I put it in SimpleMediaServer, feel free to move it.
> > Note that it has to be the Block that is cached because reconstructing
> > it from e.g. TransienStorer is currently too time-consuming.
> 
> OK, understood-- so MultiplexingMediaserver cannot do the job. I'll
> still move it out, because when using Multiplexing that can spare us
> lookup time: say you multiplex requests to one local and two distant
> mediaservers, and a specific block is available _only in the second
> distant_ mediaserver. Then caching it in that SimpleMediaserver would
> mean doing a HTTP request to the first distant mediaserver each time
> that block is looked up.

How about simply a CachingMediaserver to put on the protocol stack?

	Tuomas