platform independent kdeinit/klauncher (was KDE4's IPC)
kuba at mareimbrium.org
Tue Jan 3 13:59:36 GMT 2006
> > > > Especially for HTTP we are supposed to have a cache, so *nothing*
> > > > should be get two times.
> > >
> > > ?? This has nothing to do with the cache. The cache is only once you
> > > have fully downloaded something. But here we're talking about the very
> > > beginning of the download; you don't want to wait for minicli to
> > > download the whole thing before firing up the application [which might
> > > not even use the cache then, depending on the refresh headers and the
> > > cache policy etc.]
> > But HTTP's GET is about downloading, isn't it? Or can you tell the server
> > something like "Ooh, sorry, I do not need the rest of the information"?
> HTTP's GET does support resumption using the Range header ; so you can tell
> the server I only need so many bytes of this information or only send me
> the data from such and such bytes and up...
For data that's generated on the fly, this becomes wasteful. Think about a
simple .php page that was declared to be noncacheable. AFAIK, no server will
resume such a transfer -- it was generated on the fly, and once the
connection is closed the data is gone. I.e. you'll need to regenerate and
download the whole thing again, not just a part of it.
I think that passing the ioslave around was a good thing to do. If I were a
web application developer/server admin, I'd be needlessly worried if I saw
double GETs on some "start" pages that people are likely to type directly
into minicli. I'd suspect something wrong with the infrastructure. I'd
probably waste some time on tracking it down to KDE (mis)behaviour.
More information about the kde-core-devel