September 2009 - Dillo-dev - lists.mailman3.com

experimental patch: image loading options "same domain", "same host"

by corvid＠lavabit.com

Johannes wrote: > On Tue, Sep 15, 2009 at 07:21:32PM +0000, corvid wrote: > > Johannes wrote: > > > Hm... With "same_host" many sites no longer work because they seem > > > to use redirects (e.g. heise.de -> www.heise.de) > > > And once I have entered heise.de, www.heise.de doesn't work either. > > > > I have also found same_host not to have any value. > > > > > Generally, is it really a problem if we load url's from other > > > hosts/domains? > > > > Sites have no right to redirect me to unrelated sites, and > > sites have no right to subject me to images from unrelated sites. > > I tend to agree. With your patch I noticed that many popular sites > contain 1x1 images from some statistics gathering companies. > However I think these images do not really leak additional information as > the sites could report your IP address behind the scenes as well. I think the term "user agent" for a browser is very apt. These companies ask dillo to serve as an accomplice as they violate the user's right to privacy.

Sept. 17, 2009

experimental patch: image loading options "same domain", "same host"

by Johannes.Hofmann＠gmx.de

On Tue, Sep 15, 2009 at 07:21:32PM +0000, corvid wrote: > Johannes wrote: > > Hm... With "same_host" many sites no longer work because they seem > > to use redirects (e.g. heise.de -> www.heise.de) > > And once I have entered heise.de, www.heise.de doesn't work either. > > I have also found same_host not to have any value. > > > Generally, is it really a problem if we load url's from other > > hosts/domains? > > Sites have no right to redirect me to unrelated sites, and > sites have no right to subject me to images from unrelated sites. I tend to agree. With your patch I noticed that many popular sites contain 1x1 images from some statistics gathering companies. However I think these images do not really leak additional information as the sites could report your IP address behind the scenes as well. Cheers, Johannes

Sept. 17, 2009

View page source suggestions

by jcid＠dillo.org

On Wed, Sep 16, 2009 at 03:48:06PM +0000, corvid wrote: > Dennis wrote: > > Currently, the "View page source" right-click context menu isn't > > available below the rendered html -- so if the actual html page is very > > small and/or dillo's window is very big, right clicking the bottom half > > (for example) of the browser does nothing :\. > > This one happens to have some complexity. > > The problem is that there aren't any widgets down there and they are > what send the signals to trigger the menus. Maybe it would be possible > to increase a layout's top-level widget's size to fill a viewport > (if the layout isn't inside a button [if there's a way to check that] > and the widget is a textblock), or maybe that would break something. Yes. Maybe adding an empty widget at the bottom may solve this. > > Also, is there a way to have the source viewed in one's preferred text > > editor? :b (read: gvim) > > Only the usual way :) The idea for "View source" is to handle it with a dpi so we we can reuse dillo (it's faster and has a search function). Once this is done (which needs a review of dpip), it would be simple to adapt the dpi program to use an external editor (because a dpi program is an independent process, that in this case, would receive the source contents). I'm currently working deep in the dpi-framework. Hopefully some good news may come at the end of the month... -- Cheers Jorge.-

Sept. 16, 2009

View page source suggestions

by corvid＠lavabit.com

Dennis wrote: > Currently, the "View page source" right-click context menu isn't > available below the rendered html -- so if the actual html page is very > small and/or dillo's window is very big, right clicking the bottom half > (for example) of the browser does nothing :\. This one happens to have some complexity. The problem is that there aren't any widgets down there and they are what send the signals to trigger the menus. Maybe it would be possible to increase a layout's top-level widget's size to fill a viewport (if the layout isn't inside a button [if there's a way to check that] and the widget is a textblock), or maybe that would break something. (I actually wonder whether the signaling could be routed back through the layout or something [perhaps making handling the case easier from fltkviewbase], since all of the connectLink()s always seemed redundant to me. That might go against Dw philosophy, though, whatever that might be exactly.) > Also, is there a way to have the source viewed in one's preferred text > editor? :b (read: gvim) Only the usual way :)

Sept. 16, 2009

View page source suggestions

by dennisn＠dennisn.dyndns.org

Currently, the "View page source" right-click context menu isn't available below the rendered html -- so if the actual html page is very small and/or dillo's window is very big, right clicking the bottom half (for example) of the browser does nothing :\. Also, is there a way to have the source viewed in one's preferred text editor? :b (read: gvim)

Sept. 16, 2009

experimental patch: image loading options "same domain", "same host"

by corvid＠lavabit.com

Johannes wrote: > Hm... With "same_host" many sites no longer work because they seem > to use redirects (e.g. heise.de -> www.heise.de) > And once I have entered heise.de, www.heise.de doesn't work either. I have also found same_host not to have any value. > Generally, is it really a problem if we load url's from other > hosts/domains? Sites have no right to redirect me to unrelated sites, and sites have no right to subject me to images from unrelated sites. > Or would it be enough to not send cookies or HTTP authentication > data when loading such "unsafe" urls?

Sept. 15, 2009

experimental patch: image loading options "same domain", "same host"

by Johannes.Hofmann＠gmx.de

On Tue, Sep 15, 2009 at 08:08:06PM +0200, Joerg Sonnenberger wrote: > On Tue, Sep 15, 2009 at 07:36:31PM +0200, Johannes Hofmann wrote: > > Or would it be enough to not send cookies or HTTP authentication > > data when loading such "unsafe" urls? > > Cookies have an explicit domain part if they are supposed to apply to > more than the current host. HTTP authentication is URL specific or at > most host specific. Not sure if that helps here. If the user has a cookie set that authenticates him at www.hisbank.com and a page on www.badguy.com contains an image with the url http://www.hisbank.com?transfer_to_badguy=100000 the browser would happily send the cookie to www.hisbank.com to authenticate him. Of course www.hisbank.com should have additional checks installed that prevent such misuse. Cheers, Johannes

Sept. 15, 2009

experimental patch: image loading options "same domain", "same host"

by joerg.sonnenberger＠web.de

On Tue, Sep 15, 2009 at 07:36:31PM +0200, Johannes Hofmann wrote: > Or would it be enough to not send cookies or HTTP authentication > data when loading such "unsafe" urls? Cookies have an explicit domain part if they are supposed to apply to more than the current host. HTTP authentication is URL specific or at most host specific. Joerg

Sept. 15, 2009

experimental patch: image loading options "same domain", "same host"

by Johannes.Hofmann＠gmx.de

On Tue, Sep 15, 2009 at 07:54:44PM +0200, Justus Winter wrote: > -----BEGIN PGP SIGNED MESSAGE----- > Hash: SHA1 > > Johannes Hofmann wrote: > > Hello, > > > > On Sat, Sep 12, 2009 at 02:08:48AM +0000, corvid wrote: > >> Jorge wrote: > >>> On Sun, Sep 06, 2009 at 11:17:25PM +0000, corvid wrote: > >>>> I just put together a somewhat hacky patch to > >>>> add "same host" and "same domain" for image loading > >>>> from the tools button (or dillorc) > >>>> http://www.dillo.org/test/img_loading_options.patch > >>>> > >>>> I ran on it as long as I could tolerate images over > >>>> dialup (not long :), so here it is, and I will clean > >>>> it up if there is interest shown in getting it into > >>>> dillo. > >>> Good! I've had a similar idea for a long time. > >>> > >>> Since before the url filtering thread, I believe a simple > >>> blocking scheme based on same host/domain cuts most of the > >>> problem (bandwidth, privacy, tracking, web bugs etc). > >>> > >>> I'd suggest to add the filtering function in Capi. This is > >>> to make a_Capi_open_url() call a check function, e.g. > >>> > >>> if (Capi_filters_check_url(...) < 0) > >>> /* abort */ > >>> > >>> (you may add a ParentUrl DilloUrl to the webstructure). > >>> > >>> The point is to be able to reuse and improve the filter. > >>> (e.g. for CSS urls, sound streams, etc). > >>> > >>> > >>>> (I would probably want to change the names of the > >>>> dillorc choices for image loading, make the internal > >>>> value into an enum, look at the loadImages/autoload_image > >>>> code duplication, etc., etc... Oh, and look up how > >>>> to fix the tools menu width...) > >>> Something like: > >>> > >>> filter_remote_resources = {LOAD_ALL, SAME_HOST, SAME_DOMAIN} > >>> > >>> should be enough to get started. > >> I made a new one: > >> http://www.dillo.org/test/filtering.1.patch > >> > >> It doesn't have the menu stuff, so you have to > >> set the preference. > > > > Hm... With "same_host" many sites no longer work because they seem > > to use redirects (e.g. heise.de -> www.heise.de) > > And once I have entered heise.de, www.heise.de doesn't work either. > > > > Generally, is it really a problem if we load url's from other > > hosts/domains? > > > > Or would it be enough to not send cookies or HTTP authentication > > data when loading such "unsafe" urls? > If you're trying to protect dillo users from {X,C}SRF attacks then > stripping http authentication and cookies is a first step but it won't > protect users of intranet applications that rely on the users ip address > for authentication (think home routers with web interface). Ouch. Is there some general agreement on what web browsers should do in this regard?

Sept. 15, 2009

experimental patch: image loading options "same domain", "same host"

by 4winter＠informatik.uni-hamburg.de

-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 Johannes Hofmann wrote: > Hello, > > On Sat, Sep 12, 2009 at 02:08:48AM +0000, corvid wrote: >> Jorge wrote: >>> On Sun, Sep 06, 2009 at 11:17:25PM +0000, corvid wrote: >>>> I just put together a somewhat hacky patch to >>>> add "same host" and "same domain" for image loading >>>> from the tools button (or dillorc) >>>> http://www.dillo.org/test/img_loading_options.patch >>>> >>>> I ran on it as long as I could tolerate images over >>>> dialup (not long :), so here it is, and I will clean >>>> it up if there is interest shown in getting it into >>>> dillo. >>> Good! I've had a similar idea for a long time. >>> >>> Since before the url filtering thread, I believe a simple >>> blocking scheme based on same host/domain cuts most of the >>> problem (bandwidth, privacy, tracking, web bugs etc). >>> >>> I'd suggest to add the filtering function in Capi. This is >>> to make a_Capi_open_url() call a check function, e.g. >>> >>> if (Capi_filters_check_url(...) < 0) >>> /* abort */ >>> >>> (you may add a ParentUrl DilloUrl to the webstructure). >>> >>> The point is to be able to reuse and improve the filter. >>> (e.g. for CSS urls, sound streams, etc). >>> >>> >>>> (I would probably want to change the names of the >>>> dillorc choices for image loading, make the internal >>>> value into an enum, look at the loadImages/autoload_image >>>> code duplication, etc., etc... Oh, and look up how >>>> to fix the tools menu width...) >>> Something like: >>> >>> filter_remote_resources = {LOAD_ALL, SAME_HOST, SAME_DOMAIN} >>> >>> should be enough to get started. >> I made a new one: >> http://www.dillo.org/test/filtering.1.patch >> >> It doesn't have the menu stuff, so you have to >> set the preference. > > Hm... With "same_host" many sites no longer work because they seem > to use redirects (e.g. heise.de -> www.heise.de) > And once I have entered heise.de, www.heise.de doesn't work either. > > Generally, is it really a problem if we load url's from other > hosts/domains? > > Or would it be enough to not send cookies or HTTP authentication > data when loading such "unsafe" urls? If you're trying to protect dillo users from {X,C}SRF attacks then stripping http authentication and cookies is a first step but it won't protect users of intranet applications that rely on the users ip address for authentication (think home routers with web interface). Justus -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.9 (GNU/Linux) Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org iEYEARECAAYFAkqv1OQACgkQoPmwNWhsaZb+FACghNtaYfxcboMp1ovLmBZoy7LK MW4An08UfIfzcU2ktLqtmTLmfivzrECR =NrPe -----END PGP SIGNATURE-----

Sept. 15, 2009