- Dillo-dev - lists.mailman3.com

[patch]: Basic Authentication (updated)

by corvid＠lavabit.com

Jeremy wrote: > + > + token_value_pairs_found = 1; > + > + if (strncasecmp("realm",token,token_size) == 0 && realm == NULL) { Token could be a prefix of "realm".

Dec. 5, 2008

Dillo and Gmail

by francis＠daoine.org

On Thu, Dec 04, 2008 at 10:17:00PM -0200, Daniel Victoria wrote: Hi there, > The only thing I'm missing is Gmail. Does Dillo work with gmail? Do I > need to do any special tricks? I have it "working" with dillo1, and don't see why it shouldn't work in the same way with dillo2. "Working" means "I can read and write mail, http only" -- but only after I've got the authentication cookie with a different browser. (My dillo1 doesn't do https.) If that's good enough for you, then you can try the same with dillo2 and report on whether it still works. My cookiesrc says: DEFAULT DENY mail.google.com ACCEPT I use a mozilla browser to sign in, and then I don't sign out so as not to invalidate the cookie on the server. It seems to remain valid for a couple of weeks, so I occasionally have to redo this part. Find the cookie called GX, and write it to a file in the correct format -- 7 tab-separated fields. My current one is a file ~/gmcookie which looks like this (apart from the value of the cookie, which has been modified): mail.google.com TRUE /mail FALSE 2000000000 GX DQAAAIIAAACY9ABCDEFABCDEFABCDEFABCDEFABCDEFABCDEFABCDEFABCDEFABCDEFABCDEFABCDEFHELLOKITTYEFABCDEFABCDEFABCDEFABCDEFABCDEFABCDEFABCDEFABCDEFABCDEFABCDEFABCDEFABCDEFABCDEFABCDEFABCDEFABCDEFABCDEF__p_abcdef Then, to load or reload the cookie into dillo, I do pkill -HUP cookie pkill cookie cat $HOME/gmcookie >> $HOME/.dillo/cookies.txt And then I go to my bookmarked gmail page at http://mail.google.com/mail/?ui=html&zy=n and it all Just Works. If I get redirected to a https page, that indicates that the cookie is absent or invalid, so I put it back or get a fresh one. No guarantees that it will keep working, of course. The gmail folk can change their authentication requirements at any stage. And you miss out on the javascript goodness that is a large part of the gmail experience. And there was a stage when the html returned was broken in a way that dillo couldn't usefully handle, although that has since been changed back. But for straightforward mail tasks, it works well enough for me. Couple it with a lightweight gmail-notifier script, and you don't have to hit reload to see the new news. Good luck, f -- Francis Daly francis(a)daoine.org

Dec. 5, 2008

[patch]: Basic Authentication (updated)

by onepoint＠starurchin.org

New version, slightly amended as per previous comments. Changes: * Fixed a memory leak in Realm_add_path(). * Case-insensitive comparison in Host_from_url(). * Moved the Auth header to between Accept-Encoding and Host in a_Http_make_query_str(). Please comment, Jeremy Henty

Dec. 5, 2008

Dillo and Gmail

by corvid＠lavabit.com

Daniel wrote: > Does Dillo work with gmail? I haven't heard of it working for anyone.

Dec. 5, 2008

[patch]: Basic Authentication

by corvid＠lavabit.com

Jeremy wrote: > On Thu, Dec 04, 2008 at 05:59:21PM +0000, corvid wrote: > > > +/* > > > + * Return the host that contains a URL, or NULL if there is no such host. > > > + */ > > > +static AuthHost *Host_from_url(const DilloUrl *url) > > > +{ > > > + AuthHost *host; > > > + int i; > > > + > > > + for (i = 0; (host = dList_nth_data(auth_hosts, i)); i++) > > > + if (((strcmp(URL_SCHEME(url), host->scheme) == 0) && > > > + (strcmp(URL_AUTHORITY(url), host->authority) == 0))) > > > + return host; > > > + > > > + return NULL; > > > +} > > > > dStrcasecmp > > When comparing the scheme, the authority, or both? I think both. *pokes through rfc2396 for a bit* "When a scheme uses elements of the common syntax, it will also use the common syntax equivalence rules, namely that the scheme and hostname are case insensitive..."

Dec. 5, 2008

Dillo and Gmail

by daniel.victoria＠gmail.com

Hi all, First off, I got to say, congrats on the great browser. I have an old laptop and firefox just takes ages to load. Dillo, on the other hand, is lightning fast! Great! The only thing I'm missing is Gmail. Does Dillo work with gmail? Do I need to do any special tricks? I compiled the latest DIllo2 from svn and I did enabled https so I can get past the login page but then I'm stuck at a warning page about a non-standard redirect and, the link on that page does not work. The link is: https://www.google.com/accounts/CheckCookie?continue=http%3A%2F%2Fmail.goog… Thanks for the great work Daniel

Dec. 5, 2008

[patch]: Basic Authentication

by onepoint＠starurchin.org

On Thu, Dec 04, 2008 at 05:59:21PM +0000, corvid wrote: > Remember that there's a header ordering convention. D'oh, I saved the link to the discussion of this and then forget to review it! Will fix. > > +/* > > + * Return the host that contains a URL, or NULL if there is no such host. > > + */ > > +static AuthHost *Host_from_url(const DilloUrl *url) > > +{ > > + AuthHost *host; > > + int i; > > + > > + for (i = 0; (host = dList_nth_data(auth_hosts, i)); i++) > > + if (((strcmp(URL_SCHEME(url), host->scheme) == 0) && > > + (strcmp(URL_AUTHORITY(url), host->authority) == 0))) > > + return host; > > + > > + return NULL; > > +} > > dStrcasecmp When comparing the scheme, the authority, or both? > > +static void Realm_add_path(AuthRealm *realm, const char *path) > > +{ > > + int len, i; > > + char *realm_path; > > + > > + len = strlen(path); > > + /* ignore a trailing '/' */ > > + if (len && path[len - 1] == '/') > > + len--; > > + > > + /* delete existing paths that are inside the new one */ > > + for (i = 0; (realm_path = dList_nth_data(realm->paths, i)); i++) { > > + if (path_is_inside(realm_path, path, len)) { > > + dList_remove_fast(realm->paths, realm_path); > > + i--; /* reconsider this slot */ > > + } > > + } > > + > > + dList_append(realm->paths, dStrndup(path, len)); > > +} > > I think you need a dFree in there. For the realm_path that gets passed to dList_remove_fast()? Will fix. > PS I took the liberty of adding an entry for Basic auth to Plans.html No problem. Thanks for looking at this. Jeremy Henty

Dec. 4, 2008

[patch]: Basic Authentication

by corvid＠lavabit.com

Jeremy wrote: > ... I haven't looked into how anything fits together yet at all, but skimming through: > @@ -234,12 +236,13 @@ Dstr *a_Http_make_query_str(const DilloUrl *url, bool_t use_proxy) > "User-Agent: Dillo/%s\r\n" > "Content-Length: %ld\r\n" > "Content-Type: %s\r\n" > - "%s" > + "%s" /* cookies */ > + "%s" /* auth */ > "\r\n", > full_path->str, URL_AUTHORITY(url), > proxy_auth->str, referer, VERSION, > URL_DATA(url)->len, content_type->str, > - cookies); > + cookies, auth ? auth : ""); > dStr_append_l(query, URL_DATA(url)->str, URL_DATA(url)->len); > dStr_free(content_type, TRUE); > } else { Remember that there's a header ordering convention. > +/* > + * Return the host that contains a URL, or NULL if there is no such host. > + */ > +static AuthHost *Host_from_url(const DilloUrl *url) > +{ > + AuthHost *host; > + int i; > + > + for (i = 0; (host = dList_nth_data(auth_hosts, i)); i++) > + if (((strcmp(URL_SCHEME(url), host->scheme) == 0) && > + (strcmp(URL_AUTHORITY(url), host->authority) == 0))) > + return host; > + > + return NULL; > +} dStrcasecmp > +static void Realm_add_path(AuthRealm *realm, const char *path) > +{ > + int len, i; > + char *realm_path; > + > + len = strlen(path); > + /* ignore a trailing '/' */ > + if (len && path[len - 1] == '/') > + len--; > + > + /* delete existing paths that are inside the new one */ > + for (i = 0; (realm_path = dList_nth_data(realm->paths, i)); i++) { > + if (path_is_inside(realm_path, path, len)) { > + dList_remove_fast(realm->paths, realm_path); > + i--; /* reconsider this slot */ > + } > + } > + > + dList_append(realm->paths, dStrndup(path, len)); > +} I think you need a dFree in there. PS I took the liberty of adding an entry for Basic auth to Plans.html

Dec. 4, 2008

[patch]: Basic Authentication

by onepoint＠starurchin.org

On Thu, Dec 04, 2008 at 01:36:39PM +0100, Joerg Sonnenberger wrote: > Have you thought about how other protocols can be integrated with > this? No, except that I've deliberately made the User/Password dialog generic so that any other code can reuse it with a different callback. Other than that I think that all of my code is entirely specific to HTTP Basic authentication. I can't see how other protocols could be integrated with it in any useful sense. But my opinion doesn't count for much since I know next to nothing about how those other protocols work! > Examples are: > - TLS with client side authentication or verification for the server > certificate I have no idea how this works. Every time I try to study certificates and authentication my head starts spinning and I have to lie down. Does anyone know of some gentle introductions? > - non-anonymous FTP The FTP DPI could check the output of wget. If wget complained that the FTP server needed authentication than it could pop up the user/password dialog with a callback that retries the wget. The download DPI would have to be modified to receive the user and password. (Or we could stuff them into the download URL, but that's insecure since the password would appear in the process listing.) We would have to extend the DPI API to let DPIs use the user/password dialog but that is straightforward (I already implemented it when I thought authentication was going to be a DPI). Regards, Jeremy Henty

Dec. 4, 2008

css progress

by Johannes.Hofmann＠gmx.de

On Thu, Dec 04, 2008 at 08:13:49AM +0900, furaisanjin wrote: > Hi, > > > For now you could set force_my_colors=YES in dillorc. > > That wold be one way to avoid the problem but another issue is we > can't change dillo config on the fly, so I have to start dillo up if I > want to do so. If I keep force_my_colors=YES, dillo appearance looks > bad. That's true. I just checked your example page and it seems to look better with the current CSS prototype. The letters are visible - but still not readable for me ;-) Regards, Johannes

Dec. 4, 2008