93348 - Disabling sites from being cached

Status: RESOLVED WONTFIX

(Core :: Networking, enhancement, P5)

Description

[Marcin Orlowski]

I'd be nice to have an option to disable certain domains from being cached
(both on disk and in ram cache). it'd be very handy when you develop a site
and often test it (the cache may not be handy), or you don't need to cache
the site because it's next hop to you so there's no need to waste a disk
space for files you can fetch almost instantly.

Comment 1

[Boris Zbarsky [:bzbarsky]]

If you're developing, just have the server send back nocache headers....

Comment 2

[Marcin Orlowski]

Well, if you find playing with the server configs easier that reconfiguring the
browser, then I got nothing to say. Not to mention all the cases when you simply
can't touch the server.

Comment 3

[gordon]

The cache is used for reasons other than just performance.  It supports Session
History to allow users to (try to) redisplay a page as it was.  We want View
Source, Send Page, and Printing to utilize the cache to access the current page
as well, rather than refetching the page from the server.  There are also
multiple components that use the cache besides HTTP: FTP, IMAP, and ImgLib also
use it.  I think it would be difficult to explain to users all the side effects
of not caching certain domains.  And if you're testing a site, you could get
very different behavior than typical users, which could be the source of
additional bugs.

The cache doesn't know anything about domains; that concept belongs to the
networking protocols.  The cache simply stores data associated with a key on
behalf of its component clients (HTTP, FTP, IMAP, ImgLib, etc).  Either the
component clients would need to check a list of "no-cache" domains, or we would
need to add the concept of domains to the cache API, and perhaps fail if we got
a domain match.

For development where you can't touch the server, you could either use
shift-reload to get the latest page, or turn off caching altogether (currently
there is a bug related to this on some branches).

I'm not sure this is a widely desired enhancement, but I'd like to see a bit
more discussion before closing it.  I'm cc'ing darin to get the HTTP perspective.

 

Comment 4

[Darin Fisher]

I agree with Gordon.  You may not want to disable the entire cache, and
implementing selective caching would be a protocol specific feature.

Adding RFE to the summary... not sure which component this should belong to.
Perhaps Networking would be most appropriate.

Comment 5

[Marcin Orlowski]

I don't know the Mozilla internals, but as Gordon said, cache is protocol
specific. And all I wanted is protocos specific filter. I only want to avoid
caching http from particular domains and that seems to be doable. It should not
break history or whatelse as it already deals with the entries that are not
cacheable. So that would be kind of extension. Addditional condition to
determine if the page shall be cached or not.

Comment 6

[Darin Fisher]

File->SaveAs and View->PageSource would require a fresh download if the document
is not in the cache.  Printing does not require the HTML to be in the cache as
it can generate what it needs from the content model.

Comment 7

[Marcin Orlowski]

Woudn't it require it anyway if page is not cached due to i.e. META tag or so?

Comment 8

[Darin Fisher]

no... actually, mozilla stores just about everything in the cache unless the
server sends a 'cache-control: no-store' header.  this header is used to label
content as sensitive so as to imply that the content should not be kept around
for very long.  other cache headers, such as 'cache-control: no-cache' or
'pragma: no-cache' only prevent the browser from re-using the data for
subsequent requests (ie. link clicks).  the http/1.1 spec explicitly allows
browsers to re-use such responses for the purposes of session history
(back/forward) and the like (file->save_as, view->page_source, and printing).

Comment 9

[Marcin Orlowski]

Ok, seem we are close to the final solution ;) What about making the feature
that would simply always mark all the content from the given domains as if they
were sent with 'cache-control: no-store'? That'd give the effect I ask for.

Comment 10

[gordon]

That seems like a better approach.  Changing component.

Comment 11

[Doug Turner (:dougt)]

moving neeti's futured bugs for triaging.

Comment 12

[Daniel Wang]

*** Bug 101451 has been marked as a duplicate of this bug. ***

Comment 13

[Jon Henry]

*** Bug 217168 has been marked as a duplicate of this bug. ***

Comment 14

[benc]

-> defaults

Comment 15

[Matthew Cline]

Hardware/OS -> All/All

Comment 16

[:Gavin Sharp [email: [e-mail]]]

*** Bug 309627 has been marked as a duplicate of this bug. ***

Comment 17

[Darin Fisher]

-> default owner

Comment 18

[Patrick McManus [:mcmanus]]

dev tools has clear cache semantics