Why are web browsers (well, Netscape and its descendents) so different from all other programs?
All I want to do is run Firefox on another system on my (three machine) network and it's very difficult!
My 'main' machine runs Slackware Linux and is in the house in my office, it's on all the time so I can advertise the occasional web page etc. It's the machine that appears on the web (at least as web pages) as home.isbd.net.
I have set up another machine in the garage using tatty old hardware so when I want to get some information (thread sizes for bolts, spare parts for tools, wall plug sizes, etc.) I can do it there rather than having to go back into the house requiring cleanness etc. As I want access to the same home directory when I'm out there I have used NFS to export my home and when I log in on the garage system I have all the same files that I have on my office system.
Everything works fine *except* Firefox (well everything I've tried so far that is). Firefox won't start on the garage system because it sees the office system's Firefox lock file and says that Firefox is running already and I can't have another one. I can think of various ways around this issue (separate profile, don't use NFS but copy some files across, etc.) but really why should I have to? Just about every other program would run up a second instance without any bother at all, why does Firefox do different? Netscape was the same, it's "ancestral"! :-)
Other programs that create a lock file like this or a file with the PID in it are server type programs where it makes sense (like mySql), there having only one copy makes sense and I could use it from both my garage and office systems at the same time without hassle. With firefox though it's just a nuisance and I can't really come up with a neat solution.
To describe the problem more completely:-
I have /home/chris is the *same* file system on both office and garage computers, it's a local file system on the office computer and an NFS export to the garage computer.
I have a few files that are different for the garage computer kept in /home/chris/.garage, these are such things as my window manager configuration because the garage computer has a lower resolution display than the office computer.
I want office and garage computer firefox to share their installation and configuration so that bookmarks and other things are the same.
I just want to run firefox in the garage, but I can't!
Possible solutions seem to be:-
Run firefox in the garage with a separate profile, but I'm not sure if this works completely and it means that I have to choose a profile every time I start firefox. It also means that bookmarks etc. are *not* shared so I have to do something about that.
Shut down firefox on the office system every time I leave it, not really a sensible approach, I'll forget!
I can't think of any other ways to do it!
Oh, yes, there is another approach. Have an independent /home/chris on the garage system and have symbolic links from there to all the important directories in /home/chris on the office system. Thus firefox has a local (and thus independent) configuration/installation and I can just link bookmarks.html to the office system one to get the same bookmarks. This also has the advantage that the garage system will be more able to work without the office system being there. It's probably the approach I will take though I'm not quite sure how to keep the symbolic links up to date.
There's summat weird about the Mozilla family!
The other day I had a weird experience with them.
I had Firefox running on the laptop I was sitting at. For reasons which had to do with difficulty getting Firefox to behave on a particular website, I telnetted through to another of my 3 machines (built in 2001) and started up an old Mozilla on it (from about the same era).
Lo and Behold! I got a window spawned from the firefox on my laptop!!! No window from the Mozilla on the other machine.
This ain't the first time I've observed that these browsers seem able to detect their presence on other machines, but that was the most spectacular.
Your problems, Chris, may well be related to the same issue. Ted.
On 23-Aug-07 20:17:09, Chris G wrote:
Why are web browsers (well, Netscape and its descendents) so different from all other programs?
All I want to do is run Firefox on another system on my (three machine) network and it's very difficult! [...]
-------------------------------------------------------------------- E-Mail: (Ted Harding) ted.harding@nessie.mcc.ac.uk Fax-to-email: +44 (0)870 094 0861 Date: 23-Aug-07 Time: 22:33:16 ------------------------------ XFMail ------------------------------
On 8/23/07, Ted Harding ted.harding@nessie.mcc.ac.uk wrote:
I telnetted through to another of my 3 machines (built in 2001) and started up an old Mozilla on it (from about the same era). Lo and Behold! I got a window spawned from the firefox on my laptop!!! No window from the Mozilla on the other machine.
Did you check your DISPLAY setting? Mozilla/Firefox checks if it is already running on the current display and prods it to just open another window.
Tim.
On Fri, 24 Aug 2007, Tim Green wrote:
On 8/23/07, Ted Harding ted.harding@nessie.mcc.ac.uk wrote:
I telnetted through to another of my 3 machines (built in 2001) and started up an old Mozilla on it (from about the same era). Lo and Behold! I got a window spawned from the firefox on my laptop!!! No window from the Mozilla on the other machine.
Did you check your DISPLAY setting? Mozilla/Firefox checks if it is already running on the current display and prods it to just open another window.
And when it checks if it's already running, it checks under an application ID determined by the -a option given on the command line. So if you don't like this behaviour, you can suppress it by running the browser as
firefox -a somerubbishynamethatnoapplicationcouldhave
(Thanks are due to Chris Richardson for helping me figure that out, a few weeks back.)
The fact that this happens at all, though, means that an x client running on machine A can tell an x client running on machine B to open a new tab/window pointing at a URL chosen by the client on machine A - even though the x server may be on A or B, or indeed on a third machine C. Does anyone have a view as to the security implications of this?
On Friday 24 August 2007 11:03:49 Dan Hatton wrote:
And when it checks if it's already running, it checks under an application ID determined by the -a option given on the command line.
The fact that this happens at all, though, means that an x client running on machine A can tell an x client running on machine B to open a new tab/window pointing at a URL chosen by the client on machine A - even though the x server may be on A or B, or indeed on a third machine C. Does anyone have a view as to the security implications of this?
Excellent: another (good) reason not to use Firefox.
Konqueror ftw ;-)
Richard.
On Fri, Aug 24, 2007 at 11:25:49AM +0100, Richard Lewis wrote:
On Friday 24 August 2007 11:03:49 Dan Hatton wrote:
And when it checks if it's already running, it checks under an application ID determined by the -a option given on the command line.
The fact that this happens at all, though, means that an x client running on machine A can tell an x client running on machine B to open a new tab/window pointing at a URL chosen by the client on machine A - even though the x server may be on A or B, or indeed on a third machine C. Does anyone have a view as to the security implications of this?
Given that you've given it access to your display and handed it a shiny cookie on a plate, I don't see that as a security risk - you've granted it the permissions it requires, it's just using straight off xlib calls to check and call (IIRC).
Of course, there's nothing stopping you from taking the source of firefox and stopping this silly behaviour... while you're at it, if you can fix the memory leaks too, that'd be *great*!
Excellent: another (good) reason not to use Firefox.
Konqueror ftw ;-)
Except Konqueror sucks just as hard and uses the crappy Qt toolkit. Mix in the fact that it's not just a web browser but a nasty file manager and we appear to have gone back to around 1998 with IE4 and Windows 98... "yay"?
God I wish there was a browser that actually had: (a) decent css support (b) a fast (and accurate) rendering engine (c) no memory leaks (d) no random buffer overflows, segfaults, annoying habits (e) a UI that didn't entirely totally suck
Until that point, I'm sticking with firefox (well, actually, iceweasel in debian, at least there's half a chance that that'll get security updates)...
Ho hum - when will web browsers stop sucking and designers get the opportunity of not having to test a site it 10 million different browsers because they've all got their own quirks and interpretation of the specs - it's not like HTML is a new spec, heck, it's not even like CSS3 is new - and yet, where's the support?
Bah,
On Fri, 24 Aug 2007, Brett Parker wrote:
Given that you've given it access to your display and handed it a shiny cookie on a plate, I don't see that as a security risk - you've granted it the permissions it requires, it's just using straight off xlib calls to check and call (IIRC).
Hmm... I was fishing for someone to suggest that Xorg on machine C can be configured in a way that allows a client on machine A to open a window of its own on the display of machine C, but in which the X server on machine C doesn't pass on messages from the client on machine A that appear to be telling a client on machine B what to do. Anyone?
On Friday 24 August 2007 12:27:21 Brett Parker wrote:
Excellent: another (good) reason not to use Firefox.
Konqueror ftw ;-)
Except Konqueror sucks just as hard and uses the crappy Qt toolkit.
I've never done any serious work with Qt but it seems to have some quite cool concepts like the signal/slots thing, a nice component set, and its cross-platform.
Mix in the fact that it's not just a web browser but a nasty file manager
In fact, that's part of the appeal of Konqueror. Not so much that its a Web browser and file manager, but that its a kind of meta-application in which an KPart component can be displayed and which can make use of the kioslaves I/O abstraction layer.
I would argue that you're just biased against: Qt because of its dubious free software credentials in the past; KDE because its large and pretty; and desktop environments because you like the kudos of using the terminal for everything.
On the other hand, I use the terminal quite a bit and concede that in the hands of a proficient user it is often a more productive method of interacting with your computer especially for jobs like file management and text processing. But there are some things I really like about feature-rich desktop environments like: notification systems, clipboards (including consistent copy/paste behaviour), and homogeneous interface styling.
God I wish there was a browser that actually had: (a) decent css support (b) a fast (and accurate) rendering engine (c) no memory leaks (d) no random buffer overflows, segfaults, annoying habits (e) a UI that didn't entirely totally suck
Until that point, I'm sticking with firefox (well, actually, iceweasel in debian, at least there's half a chance that that'll get security updates)...
Ho hum - when will web browsers stop sucking and designers get the opportunity of not having to test a site it 10 million different browsers because they've all got their own quirks and interpretation of the specs - it's not like HTML is a new spec, heck, it's not even like CSS3 is new - and yet, where's the support?
Of course there is an obvious answer to this: you could make your own/improve FF.
Eeek. Please don't hurt me!
Cheers, Richard
On Fri, Aug 24, 2007 at 12:48:57PM +0100, Richard Lewis wrote:
On Friday 24 August 2007 12:27:21 Brett Parker wrote:
Excellent: another (good) reason not to use Firefox.
Konqueror ftw ;-)
Except Konqueror sucks just as hard and uses the crappy Qt toolkit.
I've never done any serious work with Qt but it seems to have some quite cool concepts like the signal/slots thing, a nice component set, and its cross-platform.
So, you're not a developer and you don't know the intracacies... well done! As for signal/slots - that's what dbus was designed for (note: don't like that crap either).
Mix in the fact that it's not just a web browser but a nasty file manager
In fact, that's part of the appeal of Konqueror. Not so much that its a Web browser and file manager, but that its a kind of meta-application in which an KPart component can be displayed and which can make use of the kioslaves I/O abstraction layer.
Errr, right.
I would argue that you're just biased against: Qt because of its dubious free software credentials in the past; KDE because its large and pretty; and desktop environments because you like the kudos of using the terminal for everything.
When did KDE get "pretty", it's been ugly as sin for as long as I can remember, and I haven't seen it looking nicer recently. Qt is a crappy toolkit but has a nice API - if it wasn't for the fact that it is so damned ugly it might be worth time investigating, but, no, it's ugly and not as portable as GTK... (note: I hate new versions of GTK with a passion too, mostly because Gnome took over development of it and then fucked it over in "new and interesting" ways, Gtk-- isn't bad for the C++ peeps out there.
On the other hand, I use the terminal quite a bit and concede that in the hands of a proficient user it is often a more productive method of interacting with your computer especially for jobs like file management and text processing. But there are some things I really like about feature-rich desktop environments like: notification systems, clipboards (including consistent copy/paste behaviour), and homogeneous interface styling.
God I wish there was a browser that actually had: (a) decent css support (b) a fast (and accurate) rendering engine (c) no memory leaks (d) no random buffer overflows, segfaults, annoying habits (e) a UI that didn't entirely totally suck
Until that point, I'm sticking with firefox (well, actually, iceweasel in debian, at least there's half a chance that that'll get security updates)...
Ho hum - when will web browsers stop sucking and designers get the opportunity of not having to test a site it 10 million different browsers because they've all got their own quirks and interpretation of the specs - it's not like HTML is a new spec, heck, it's not even like CSS3 is new - and yet, where's the support?
Of course there is an obvious answer to this: you could make your own/improve FF.
Eeek. Please don't hurt me!
It's on my to do list - along with (now) finishing off a window manager that doesn't suck and doesn't have a complete twazzock as a "maintainer"... a Free java implementation that doesn't entirely suck... a decent Free search backend (though - actually - looks like xapians example backend is actually really rather good)... a decent CMS (that doesn't leak memory, doesn't take a year and a half to do anything and doesn't annoy the hell out of me)... etc, etc, etc.
Hum ho - of course, everyone is entitled to their opinion - but *anyone* that thinks that *any* of the web browsers out there at the moment is worth anything is seriously mistaken - they all suck in their own ways - wether it's because the CSS rendering engine is *too* lenient (*glares at Konqueror*), under implemented (*glares at firefox and IE 7*) or just entirely missing (*glares at 99.9% of other browsers).
Anyways - just got back from "1 or 2" beers, and am thus going the heck to sleep.
Brett Parker iDunno@sommitrealweird.co.uk wrote:
On Fri, Aug 24, 2007 at 12:48:57PM +0100, Richard Lewis wrote: [...]
I've never done any serious work with Qt but it seems to have some quite cool concepts like the signal/slots thing, a nice component set, and its cross-platform.
So, you're not a developer and you don't know the intracacies... well done! As for signal/slots - that's what dbus was designed for (note: don't like that crap either).
Indeed, dbus is a better way to do it, but it's still a bit vexing.
Qt's preprocessor abuse was horrible horrible and if you tried to build other software which also used their own Very Special preprocessors... oh, the pain! aaargh! ;-) I really hope they've fixed that by now and the preprocessor has gone and died in a corner, but I've avoided Qt for some years now.
In fact, that's part of the appeal of Konqueror. Not so much that its a Web browser and file manager, but that its a kind of meta-application in which an KPart component can be displayed and which can make use of the kioslaves I/O abstraction layer.
Errr, right. [...]
ITYM I/O abstraction should be done in the I/O layer (as in FUSE) not each and every desktop environment.
Of course, I'm also biased against KDE because Stefan Kulow once replied to my bug report of a fork-bomb with a suggestion that I should buy more RAM to see if that solved the problem... long before I understood all the QPL/GPL/CoV controversies.
Regards,
On 25/08/07, MJ Ray mjr@phonecoop.coop wrote:
Of course, I'm also biased against KDE because Stefan Kulow once replied to my bug report of a fork-bomb with a suggestion that I should buy more RAM to see if that solved the problem... long before I understood all the QPL/GPL/CoV controversies.
LMAO!
I havd to ask ... did it?
hehe
Peter.
samwise samwise@bagshot-row.org wrote:
On 25/08/07, MJ Ray mjr@phonecoop.coop wrote:
Of course, I'm also biased against KDE because Stefan Kulow once replied to my bug report of a fork-bomb with a suggestion that I should buy more RAM to see if that solved the problem... [...]
I havd to ask ... did it?
Back in those days, we UEA students couldn't afford more RAM. We just got High Performance Computing accounts instead. Never did get KDE built on hpc1. Needed more RAM, I guess. ;-)
Regards,
On Friday 24 August 2007 12:27, Brett Parker wrote:
Except Konqueror sucks just as hard and uses the crappy Qt toolkit.
Not wishing to gainsay, but I'd be interested to hear your thoughts on what makes Qt a crappy toolkit.
From what I can see, Qt is a stable, lightweight, featureful and reasonably documented toolkit and some sound design ideas, available as Free software - I'd be interested to know how it comes off poorly against almost anything else available.
Not that it's perfect, of course - nothing is, and what foibles it has can jump out at you - I've had to hack around threading issues in the past. Sure as hell beats _porting_ apps between point releases on other toolkits into a cocked hat though. I hear a lot of superstition about how terrible Qt is, and not much reason.
I doubt you're likely to hold up a Mozilla platform or a Gtk as a shining example of toolkit brilliance, but I am curious to find out why you say this and whether you have some secret GUI toolkit in mind I don't know about.
(Also, KHTML is great, I suggest you have a nose around it at some point! :-) )
God I wish there was a browser that actually had: (a) decent css support (b) a fast (and accurate) rendering engine (c) no memory leaks (d) no random buffer overflows, segfaults, annoying habits (e) a UI that didn't entirely totally suck
It's very much these kinds of considerations that have me choosing Konqueror over Firefox for light browsing tasks, as the better of various imperfect browsers (well, on-balance. It's orders of magnitude snappier for most pages, and where I do want to fall back on any other browser for heavy js or ie quirks, there's a convenient option to in the menu).
Also, any kind of AA/SPH for fonts seems to look horrendous in firefox and lovely in konqy, which makes web documentation a little easier on the eye (for me anyway, YMMV).
Until that point, I'm sticking with firefox (well, actually, iceweasel in debian, at least there's half a chance that that'll get security updates)...
Ho hum - when will web browsers stop sucking and designers get the opportunity of not having to test a site it 10 million different browsers because they've all got their own quirks and interpretation of the specs - it's not like HTML is a new spec, heck, it's not even like CSS3 is new - and yet, where's the support?
Bah,
When the current generation of browsing bloatware dies, and browsers can track standards instead of what IE/Mozilla burp at them.
Let's not hold our breaths on that one.
On Fri, Aug 24, 2007 at 12:27:21PM +0100, Brett Parker wrote:
God I wish there was a browser that actually had: (a) decent css support (b) a fast (and accurate) rendering engine (c) no memory leaks (d) no random buffer overflows, segfaults, annoying habits (e) a UI that didn't entirely totally suck
Until that point, I'm sticking with firefox (well, actually, iceweasel in debian, at least there's half a chance that that'll get security updates)...
You still not written bpzilla-explorer.sh yet? Sheeeesh! You've even got holiday at the moment! ;)
Adam
On Fri, Aug 24, 2007 at 02:52:08PM +0100, Adam Bower wrote:
On Fri, Aug 24, 2007 at 12:27:21PM +0100, Brett Parker wrote:
God I wish there was a browser that actually had: (a) decent css support (b) a fast (and accurate) rendering engine (c) no memory leaks (d) no random buffer overflows, segfaults, annoying habits (e) a UI that didn't entirely totally suck
Until that point, I'm sticking with firefox (well, actually, iceweasel in debian, at least there's half a chance that that'll get security updates)...
You still not written bpzilla-explorer.sh yet? Sheeeesh! You've even got holiday at the moment! ;)
Not yet - been, errr, lazy for a change and mostly offline for the last week - it's been marvelous! Not going to be properly engaged with a terminal till tuesday - occasionally it's nice to have a break :)
I was half thinking of writing a browser in python - I don't think bash and standard unix utils *quite* have enough to make a decent browser yet ;) Might just about get away with writing a browser in zsh, though ;)
On Fri, Aug 24, 2007 at 11:03:49AM +0100, Dan Hatton wrote:
The fact that this happens at all, though, means that an x client running on machine A can tell an x client running on machine B to open a new tab/window pointing at a URL chosen by the client on machine A - even though the x server may be on A or B, or indeed on a third machine C. Does anyone have a view as to the security implications of this?
Yes, I knew about this aspect of Netscape/Mozilla/Firefox and it was something that used to cause me hassle at work. Linux on desktop machine and, of course, I had Firefox running there. If you then have a terminal window open on a Sun machine (our development target) and run Firefox it simply opens a new tab in the Linux one. Since the Linux firefox was often running on a different desktop it was very easy to miss it completely and think the your Firefox had been lost.
There's a way of preventing this different from what was given here, use the -no-remote command line option or set MOZ_NO_REMOTE in your environment.
However this doesn't really bear directly on my problem where I'm trying to run Firefox on a different system and a different DISPLAY, it just happens to have the same configuration files.
-
Adam Bower -
Brett Parker -
Chris G -
Dan Hatton -
MJ Ray -
Richard Lewis -
samwise -
ted.harding＠nessie.mcc.ac.uk -
Ten -
Tim Green