RE: Web browsers should preserve the file system Last-Modified time of downloaded files

RFC2616 [0] says "HTTP/1.1 servers SHOULD send Last-Modified whenever feasible", and suggests that "For files, it may be just the file system last-modified time".  However, this header is only mentioned for use in caching scenarios, and the specs are mute on UA behavior for "file save".

Apart from using file-centric protocols like FTP or WebDAV, users can opt-in to using HTTP last-modified header on download timestamps by using wget or one of the download managers that has the option.  Wikipedia's "modified timestamp preserved" column [1] doesn't seem very helpful for determining how popular this option is among users of download managers.  FWIW, bittorrent doesn't support last-modified preservation. 

[0] http://www.w3.org/Protocols/rfc2616/rfc2616-sec14.html#sec14.29
[1] http://en.wikipedia.org/wiki/Comparison_of_download_managers#Features

-----Original Message-----
From: Larry Masinter [mailto:masinter@adobe.com] 
Sent: Saturday, November 05, 2011 9:33 AM
To: Brolin Empey; www-tag@w3.org
Subject: RE: Web browsers should preserve the file system Last-Modified time of downloaded files

The web architecture at least as we have specified it so far does not address the notion of "downloading" and of "files" or modified dates.

As far as the web is concerned, the data stream you get from a GET is at least potentially reinvented every time.

And clocks often can be skewed, so that the last-modified timestamp that you get is not in your epoch anyway, so setting file mod time anything other than local time of receipt isn't useful.


-----Original Message-----
From: Brolin Empey [mailto:brolin@brolin.be] 
Sent: Friday, November 04, 2011 4:19 PM
To: www-tag@w3.org
Cc: Tim Berners-Lee; Amy van der Hiel
Subject: Web browsers should preserve the file system Last-Modified time of downloaded files

Hello,

I have already contacted timbl and amy (in CC above) about this subject because I wanted timbl's personal opinion, but timbl referred me to this list.

Most GUI Web browsers (maybe most Web browsers, including non-GUI ones?) do not preserve the file system last modified time of downloaded files: 
  even if the HTTP server sends the Last-Modified response header, the Web browser does not set the last modified time (mtime) field of the downloaded file in the local file system to the Last-Modified time served with the file.  This behaviour effectively *changes* the last modified time of the downloaded file even if the file has not actually been modified.

The Web browser should preserve the file system last modified time by default because this time cannot easily be recovered after it is discarded.

I believe implementing this basic file copy functionality should be prioritised above advanced features such as WebGL and Typed Arrays [1]:

current releases of both Mozilla Firefox [2] and Google Chrome [3] can run Quake II [4] [5] and boot and run a Linux v2.6.20 kernel and BusyBox userland on an x86 computer emulated in JavaScript [6] without requiring any extensions or plug-ins but still cannot even preserve the file system last modified time of a downloaded file without an optional extension (for Mozilla only, not Google Chrome).

Mozilla Firefox >= v5.0 can preserve the file system last modified time if the optional Preserve Download Modification Timestamp (PDMTS) extension is installed:

<https://addons.mozilla.org/en-US/firefox/addon/preserve-download-modification/>

This extension probably would not have been created if I had not acted to fix Mozilla:

<http://forums.mozillazine.org/viewtopic.php?p=10608005#p10608005>

I used to believe my priorities in life were inverted, but my priority inversion seems to have been outdone by the developers of major GUI Web browsers.

I do not want to start Yet Another(tm) debate over the last modified time preservation behaviour but, if necessary, I can fully support my argument, as I have already done elsewhere multiple times.

I want the W3C to convince the developers of (at least major GUI) Web browsers to finally fix this bug by preserving the file system last modified time of downloaded files by default.  The behaviour could be a policy with an easily discoverable GUI to change it because I know some confused users will resist change by wanting the broken behaviour.

At the least, I want Mozilla to include PDMTS in the base installation of both Firefox and SeaMonkey with file system last modified time preservation enabled by default.

Thank you,
Brolin Empey

[1] <http://www.khronos.org/registry/typedarray/specs/latest/>
[2] Tested with v5.0 on Windows 7 Professional Service Pack 1 IA-32 [3] Tested v15 on on Windows 7 Professional Service Pack 1 IA-32 [4] <http://code.google.com/p/quake2-gwt-port/>
[5] <http://playwebgl.com/games/quake-2-webgl/
[6] <http://bellard.org/jslinux/>

Received on Saturday, 5 November 2011 17:32:34 UTC