On Wed, May 02, 2001 at 08:55:19PM -0500, Brandon wrote:
> 
> > The correct behavior would be to send the compressed data along to the
> > browser without touching it, while setting the Transfer-Encoding HTTP
> > header to gzip.  If the browser is incapable of handling that
> > Transfer-Encoding then you can unzip it yourself and send it along.
> 
> That's the right way if you're writing a browser-centric metadata system.
> However, the web is only one of the possible interfaces. What you're
> proposing would break all existing clients whenever this feature is used
> just so that things would be the most natural for the one client that also
> integrates with a web browser. I see no value in doing it the way which is
> most natural for integrating with a web browser.

No, I'm saying that FProxy should do the right thing w/r/to a web browser.
Other clients may or may not need to perform automatic decompression at
layer X.  How can adding a feature to FProxy break all existing clients?

> > But how are you supposed to reliably determine the Content-Type to
> > send to the browser if we're sticking application/x-gzip in our
> > ContentType Info field?  Filename guessing is not sufficient..
> 
> Obviously if the clients are by convention automatically zipping/unzipping
> files then they will by convention store the MIME type of the zipped file
> in the generated zip file. So when the autozip-aware client sees a zip
> file it unzips it (if autounzip is turned on). If the file contains a file
> called CONTENT-TYPE then it reads the content type from it, sets the
> ContentType field in the metadata, and sends the contained file along as
> the data. If no such file exists, then it can either 1) use filename
> guessing if configured to do so, or 2) give up since this is just a normal
> zip file and not an autozipped file. Of course if filename guessing is
> turned on it will work with normal zip files as well. However, if there is
> no content type and file name guessing is off then the only reasonable
> thing to do is to give up and pass the zip on as a zip.

Don't conflate compression with archiving.

> This scheme will actually work with clients which are totally naive to
> autozipping. You can put the CONTENT-TYPE filed in a zip file yourself or
> turn on filename guessing. Or you can just download straight zip files and
> unzip them yourself.

You're right.  That is so much simpler and transparent than, say, adding
a ContentEncoding field to the Info metadata.

> > Let's do this the right way.
> 
> Definitely.

La la la.

-- 

# tavin cole
#
# "The process of scientific discovery is, in effect,
#  a continual flight from wonder."
#                                   - Albert Einstein


_______________________________________________
Devl mailing list
Devl at freenetproject.org
http://lists.freenetproject.org/mailman/listinfo/devl

Reply via email to