Re: Status of PUT?

Roger Gonzalez writes:
Thx for feedback & infos,
 > >>>>> Laurent Demailly writes:
 > Laurent> I have an implementation of an httpd using PUT (along with
 > Laurent> DELETE and a 'new' CHMOD method).
 > I also have a server that does PUT, DELETE, and the rest of the
 > lesser-used methods.  I ran into a similar need for a CHMOD-like

CHMOD name sounds maybe too close to Unix chmod(1) though It was not
in my mind, It was use saying "change mode" with "mode" being
additionnal (meta) informations, outside the pure content

 > functionality, but I handled it in a more generic manner as a
 > MODIFY method, with a couple new header fields.  MODIFY is a lot
 > like HEAD, except that it -stores- the header info.  MODIFY lets
 > you rename a resource, change its access characteristics, change
 > its description, etc.  I use the same headers to initialize the
 > characteristics of a new resource.

but ok, I won't mind to change to a 'standard' name, like using MODIFY
[though maybe modify sound it will modify the content ?]
Any other suggestions ?

 > For the record, headers that I have found useful include:
 >   X-Access: realm:flags,realm:flags,realm:flags
 >   X-Delete-after: Some Date
 > Because Expiration isn't defined to do what I need.  (I use this for
 > when the user (usually the user's application) is uploading a
 > temporary file.

I am planning to use Expire for exactly the same purpose, why Expire
(sent by the PUTing client) won't fit ?

 > Laurent> As a side note, I am using the following addition to the
 > Laurent> protocol : (is that worthing standardization ?)  new headers
 > Laurent> : "Control:" with keywords, I currently use "mode=[r-][w-]"
 > Laurent> read/delete mode which is use by clients in PUT or CHMOD to
 > Laurent> set/change modes

 > It strikes me that this would end up being a catch-all header.  I'd
Yes, but why not [not all, but all control/access/meta changes?] ? more
catch-all than X-Access ?
 > rather see if we can come up with some more generic headers that would
 > correspond to the most common keywords that you are suggesting.
I was thinking that having one 'topical' header with some optional and
expandable keywords was more efficient and cleaner that multiplying
the headers, am I completly offbase ?

 > Laurent> "Content-Checksum:" with keyword "MD5=<hexasciimd5sum>"
 > I also use an optional checksum header.
Maybe we could use the same protocol ?

 > Laurent> While I'm at it, I'd like to point out that lots of servers
 > Laurent> and client (like Mozilla on PCs and mac or CERN' server) 
 > I think that even the 2.0beta version of Netscape sends old-style
 > dates. 

Laurent Demailly * * Linux|PGP|Gnu|Tcl|...  Freedom
Prime#1: cent cinq mille cent cinq milliards cent cinq mille cent soixante sept

Peking Marxist terrorist Chirac nuclear NORAD arrangements

Received on Monday, 16 October 1995 06:38:33 UTC