Re:Random RISC OS trivia by an ex-user
on
The ROX Desktop
·
· Score: 1
> I find it astounding that the default for RPM is to put EVERYTHING in the/usr
It is reasonable if you want your shell not to search among a thousand directories when you want to launch an application. And since RPM provides an easy way to remove everything of that app, there's no need to do so.
The need arises when you compile yourself something, so "./configure --prefix=/Apps/Someapp" makes sense. But then you could as well create a package on the fly, and use that to keep track of files.
5.There needs to be optional "automagic" configurations. One-button installation, for given categories of user
And this leads to: "wich category am I in today?". See, it is useful when you bring up a whole system you never used before (and it's actually done in Linux installations), but it's absolutely not a slver bullet for everything (anyone remembers IE eating up gobs of disk space for the cache because of the concept that "10% should be enough for everyone" and you never went there to check?).
8. Documentation should be in hypertext format
Again, it's not a good idea if you want to have a good manual from your document. Instead of pure hypertext we could have well organized hierarchical structures with some hyperlynks only when absolutely needed. See the GNU documentation.
10. PPP is not as friendly as it could be
I always explain how to create manually the 3 text files needed for a dialup connection when someone asks me about a kppp/netcfg/whatever problem. It always worked well, and no one complains.
For "newbies" who are at the point of wanting to do kernel upgrades, there needs to be a script which sets the defaults to intelligent values
Precompiled kernels are just fine for newbies. Because if one compiles the kernel not knowing what he's doing, the results are BAD ("oh, I don't have a network, so I don't need networking code". *poof*).
Sure, you might, but doesn't it seem likely that you are more likely to want a kernel on the system you're using?
Not if you are preparing a kernel for OTHER systems, since is what one usually do (compile on fast machine - install on slow machine).
texinfo is nearly impossible to manage since it requires sophisticated tools that don't play nice with anything that anyone actually uses. Plus, it requires a central table of contents which is difficult to manage automatically fromun/installation scripts
If you are unhappy with TexInfo, I am not with you. From it you can easily generate ASCII and ISO Latin 1/2 Text, dvi, (and then PostScript and pdf, indirectly), HTML, groff and other formats, all with good indices and references. And, of course, the sophisticated tool you are talking about is Emacs, which offers easy means to work with TexInfo, but any text editor with a regexp search/replace function is equally valid.
Don't do the mistake of considering TexInfo and.info files the same thing (only because most distributions install only the later by default). The later are exclusively for online browsing, and I admit that it may be difficult if you use the "info" browser without knowing nothing of Emacs (info emulates an Emacs environment just for what concerns reading documentation), but ahere are several alternatives to that (i.e. see the help system embedded in KDE, or tkinfo, or servers that do on-the-fly HTML rendering which you can read then with any browser from Lynx to Navigator).
Give it a try: I'm translating the Emacs documentation in Italian (716Kb translated till today), and I still find TexInfo a good tool for the purpouse, with a nice macro language just in the case you need it. No fancy things. Only what's needed for a good printed manual and online documentation.
According to this page of the Prisma Engineering website (human translation follows):
Embedded Linux Solutions
Among the Linux-based projects developed by us there is Touchphone, a smart phone that allows to send and receive calls, fax and email, other than having the functionalities of an organizer/agenda. The user interface is aimed toward the maximum ease of use, and it's based on an LCD display with a touch screen. There isn't any keyboard other than the one displayed on the screen.
The phone is based on a relatively simple and cheap hardware platform. Linux presence is seen only at power-on and power-off; other than that, the specific graphical user interface hides the underlying complexity of the product.
Kernel robustness and the environment as a whole is proved by the hundreds of Touchphones in exercise among the final users, with an high satisfaction index. In addition, Touchphone does not require administration by the user, who is completely unaware of using a real Unix system in its phone.
> Linux uses one entire "DOS" extended partition for it's further slicing into swap and fs points.
Please, don't take me as a Linux Troll, but I really don't understand what you said: the case you cite is just a possibility out of many.
Nothing prevents you to use 2 "DOS" primaries (as you call them), one for the fs, the other for the swap space). Or just one with the swap space in a file mounted via loop device (slower, of course). There's no requirement that Linux (and swap) stays in a particular partition.
Instead, a really nice feature that FreeBSD has and Linux hasn't yet is the layering of multiple filesystems one on top of each other (that is, the result is that in the same subtree you find files/directories belonging to different filesystems, all joined together without the need to do symlinks everywhere).
USA is the place where everything is allowed, unless it's illegal.
USSR was the place where everything was forbidden, unless it was legal.
Italy is the place where everything is is allowed, expecially what is forbidden.
AFAIK, Italy is the only nation where laws distinguish between "forbidden" and "severely forbidden, as if there were different grades of "forbidden". THIS is dumb.:-) Not cars circulating only 5 days a week (which has sense if it happens to you to be a shopkeeper in a town built in the middle age waiting for deliveries)
Please note that this habit (of referring to X as the "X WindowS") had made it even in official GNU documentation (i.e. "X Windows" everywhere in the GNU Emacs 20.4 manual).
Perhaps it was done intentionally? Perhaps it's time to give it up, and add another row to the manpage?
At last, I see that using the right term is a way to show some respect to those people that provided us all with this great piece of well-engineered and well-thought software (...yes, I know that NeWS was technically better...).
> Why not just support internationalization through gettext?
Because GNU gettext is under GPL, so...:-)
But then, there is a more serious issue than that: actually, the framework they are using doesn't do any layout on dialogs, so it's likely that longer strings on labels/buttons/etc. would be displayed only partially.
Yes, it's definively a bad idea, but this is a trick to obtain more refresh speed, and AFAIK it's quite common in the Windows world, where geometry recalculation is done once, or none at all, and this is also the reason why UNIX toolkits seem slower. Actually, toolkit like GTK, or Motif, or Qt manage the layout by default (and the fixed-width/fixed position buttons are the exceptions, not the rule).
Last, StarOffice comes with a nice spellchecker, and that has to be made up for every single language (simply checking single words is definitively not enough in some languages like italian).
> False results. If the data format etc. are known, it's possible to feed the servers bogus results, which could lead to inconsistencies in the data base
Send the same data to multiple receivers (randomly chosen), and see if they produce the same results. (or, at least, choerent ones). If note, one (or possibly more) are lying. Anyway, a closed-source client does not prevent someone to see what it does and send bogus data anyway. It only makes things harder for the ones that actually want to send correctd data.
> Data Theft. An open source program could be modified by Big Bad Corporation Inc. to simply harvest raw data and feed it into their own computers
This is a more realistic issue, but Big Bad Corporation is probably rich enough to do reverse engineering of the protocol by itself, and access random lumps of raw data anyway. A closed-source client don't make much sense here.
The real point is that modified versions (i.e. to improve performance) could quickly spread so that just a few uses the original clients.
If suddenly it turns out that a widespread modified version produces erroneous data from time to time, then probably large amount of computations has to be thrown away. Of course, you could check for that using the same method you use to check for "bad guys", but it's a serious problem if you got only few people running the original.
My 0.02 Euro as usual.
Re:RH 6.1 - NS Home Page IS Set To Redhat.com!
on
redhat.com Redone
·
· Score: 1
> It was done in the/usr/bin/netscape script as a forced command line parameter when the browser is opened without any other parameters
Well, SuSE also does this, but only if you didn't set up a homepage (that is, the wrapper checks for preferences.js and act accordingly).
> GraphOn produces networking products, not Windows emulation. This is basically a Windows version of X-Windows
From what I understand reading on the GraphOn website, it seems to me that their product does exactly the job done by the well-known (and GPL'd) VNC, perhaps better.
> Windows by default uses this kind of approach, no need to use "extension"
FYI, at the time X11 was first released (1987), designers were broaded-mind enough to leave room for extensions of the protocol and of transport means. Today, SHM support is a de-facto standard on X servers (both commercial and XFree86), but it is still called 'extension' with respect to the protocol as defined in 1987.
Btw, I'd suggest you to not underestimate the capability of working over networks. That is exactly what many companies are wanting to sell as an innovation (see Windows Terminal server, Metaframe, PC/Anywhere, thin clients and so on), and X protocol is definitively faster (because only primitives has to be transfered, not bitmaps) on this side. Have a nice day.
> MS is getting to be less and les a monopoly, because people are voting with their dollars
Although the scenery you prospect is really desiderable, please stop a moment to think about who has money to spend, and who has the power to decide where it should be spent.
In a perfect world, this wouldn't be a problem. In real world, there are people that have no choice but to accept what the market gives to them, because they may afford only the cheapest things. So, the cheapest wins.
Don't you think that a judgement such that was possibile because millions decided to give away their precious software for free with respect to freedom, breaking the market well-established rules?
> I'm not sure that you can distribute kernel modules without some kind of GPL
As an exception Linus made, you may distribute binary-only modules, and the kernel has not to be modified in any way. But then the maintenance is completely up to you. This is how it has been done for the SBLive drivers, until Creative released the sources.
Please note that modules binaries are strongly kernel-version dependant, so if you provide a binary-only module, you'd provide it at least for more than one version of the kernel. This usually is bad (because you are stick with a few kernel versions supported), but IMHO it's not so bad in this case (a firewall machine is a firewall and nothing else, usually).
> As I understand it, VNC is less like X and more of a constant-screengrab thing
Yes and no.
The server part on UNIX machines is just like another X server, only it pipes data through the net instead of video). Of course, you can have as many as you want running on a UNIX system (provided you have enough resources).
OTOH, on Windows and MacOS, it grabs the whole screen, so it may be useful only as a remote administering tool (ala Pc Anywhere).
The protocol being used actually transfers images in a way which seems to be quite efficient (but far from the efficiency obtained transfering mostly primitives). Still, it's really usable on a LAN.
There are platform-specific clients, and there's an implementation as a Java applet that just runs fine in a web browser. The interesting thing is that the VNC server part provides also an http interface that lets a user with only a web browser to authenticate and download&execute the Java client on the fly.
> I'm sure emacs could be programmed to be a full featured IDE
I don't really want to start another holy war here, but for me (and ERS too:-) it's already a full-featured IDE, expecially when you use the OO Browser from BeOpen. Surely, it takes some time to learn to use it efficiently. It does not slam on your face every feature it has, it may seem that it behaves strangely compared to other similar products, but don't forget that it comes out from over 20 years of experience and improvements.
Oh, yeah, I don't mind if Linus considers it "pure evil":-)
> shareware like WinZip and the like might get ported... and I'm sure someone at WinAmp would take a swat at it...
Bad examples: both exists already as native free application on Linux (well, actually not clones, but they provide the same user interface and the same functionalities). Probabily, there would be little or no commercial interests at all in porting applications like these, because the free ones are most of the times already there. Let's call it "natural software selection": the better ones are used at large, the worst ones simply die, or are used only by a restricted number of people that have a good reason to use them. Surely, not an interesting target for shareware makers (which usually need a large user base to be motivated, or at least the perspective of having a large paying user base, otherwise they'd used a free license).
But then, was the win32 API designed with portability in mind? Is there actually the will to make/keep it portable? I don't think so. That's seems to me just a temporary patch, not a real solution.
> Stats I have seen (but have no pointers to) say CORBA is twice as popular as DCOM.
Surely, all telco companies use CORBA-based distributed systems. AFAIK, the rare times a DCOM interface is offered in these circumnstaces is just a 'plus'.
And then, there is already a standard protocol (IOOP) and a standard encapsulation in http (IOOP on HTTP).
So are we going to get also scrollbars that see in advance which part of those _damned_ fixed width dialog windows the user wants to see? Scrollbars are elevators, after all:-)
At least, it would be useful. But no, I wouldn't like scrollbars going on strike and feel depressed because they only go up and down (or left and right).
> Linux supplements HP-UX, it doesn't/shouldn't/will never replace HP-UX
Premise: I use both Linux and HP_UX, and I like both of them. Evidently, Sanyo does not think this way.
Excerpts from the article above:
Sanyo, a $17 billion consumer electronics manufacturer, will be using TurboLinux as the base operating system in 20,000 Newve medical workstation products it expects to ship over the next four years.
...
The workstations replace PA-RISC Hewlett-Packard workstations running HP (UX).
Slashdot Mirror
> I find it astounding that the default for RPM is to put EVERYTHING in the /usr
It is reasonable if you want your shell not to search among a thousand directories when you want to launch an application. And since RPM provides an easy way to remove everything of that app, there's no need to do so.
The need arises when you compile yourself something, so "./configure --prefix=/Apps/Someapp" makes sense. But then you could as well create a package on the fly, and use that to keep track of files.
5.There needs to be optional "automagic" configurations. One-button installation, for given categories of user
And this leads to: "wich category am I in today?". See, it is useful when you bring up a whole system you never used before (and it's actually done in Linux installations), but it's absolutely not a slver bullet for everything (anyone remembers IE eating up gobs of disk space for the cache because of the concept that "10% should be enough for everyone" and you never went there to check?).
8. Documentation should be in hypertext format
Again, it's not a good idea if you want to have a good manual from your document. Instead of pure hypertext we could have well organized hierarchical structures with some hyperlynks only when absolutely needed. See the GNU documentation.
10. PPP is not as friendly as it could be
I always explain how to create manually the 3 text files needed for a dialup connection when someone asks me about a kppp/netcfg/whatever problem. It always worked well, and no one complains.
For "newbies" who are at the point of wanting to do kernel upgrades, there needs to be a script which sets the defaults to intelligent values
Precompiled kernels are just fine for newbies. Because if one compiles the kernel not knowing what he's doing, the results are BAD ("oh, I don't have a network, so I don't need networking code". *poof*).
Sure, you might, but doesn't it seem likely that you are more likely to want a kernel on the system you're using?
Not if you are preparing a kernel for OTHER systems, since is what one usually do (compile on fast machine - install on slow machine).
texinfo is nearly impossible to manage since it requires sophisticated tools that don't play nice with anything that anyone actually uses. Plus, it requires a central table of contents which is difficult to manage automatically fromun/installation scripts
.info files the same thing (only because most distributions install only the later by default). The later are exclusively for online browsing, and I admit that it may be difficult if you use the "info" browser without knowing nothing of Emacs (info emulates an Emacs environment just for what concerns reading documentation), but ahere are several alternatives to that (i.e. see the help system embedded in KDE, or tkinfo, or servers that do on-the-fly HTML rendering which you can read then with any browser from Lynx to Navigator).
If you are unhappy with TexInfo, I am not with you. From it you can easily generate ASCII and ISO Latin 1/2 Text, dvi, (and then PostScript and pdf, indirectly), HTML, groff and other formats, all with good indices and references. And, of course, the sophisticated tool you are talking about is Emacs, which offers easy means to work with TexInfo, but any text editor with a regexp search/replace function is equally valid.
Don't do the mistake of considering TexInfo and
Give it a try: I'm translating the Emacs documentation in Italian (716Kb translated till today), and I still find TexInfo a good tool for the purpouse, with a nice macro language just in the case you need it. No fancy things. Only what's needed for a good printed manual and online documentation.
According to this page of the Prisma Engineering website (human translation follows):
Embedded Linux Solutions
Among the Linux-based projects developed by us there is Touchphone, a smart phone that allows to send and receive calls, fax and email, other than having the functionalities of an organizer/agenda. The user interface is aimed toward the maximum ease of use, and it's based on an LCD display with a touch screen. There isn't any keyboard other than the one displayed on the screen.
The phone is based on a relatively simple and cheap hardware platform. Linux presence is seen only at power-on and power-off; other than that, the specific graphical user interface hides the underlying complexity of the product.
Kernel robustness and the environment as a whole is proved by the hundreds of Touchphones in exercise among the final users, with an high satisfaction index. In addition, Touchphone does not require administration by the user, who is completely unaware of using a real Unix system in its phone.
It's not clear.
> Linux uses one entire "DOS" extended partition for it's further slicing into swap and fs points.
Please, don't take me as a Linux Troll, but I really don't understand what you said: the case you cite is just a possibility out of many.
Nothing prevents you to use 2 "DOS" primaries (as you call them), one for the fs, the other for the swap space). Or just one with the swap space in a file mounted via loop device (slower, of course). There's no requirement that Linux (and swap) stays in a particular partition.
Instead, a really nice feature that FreeBSD has and Linux hasn't yet is the layering of multiple filesystems one on top of each other (that is, the result is that in the same subtree you find files/directories belonging to different filesystems, all joined together without the need to do symlinks everywhere).
Just a little note about Italy:
:-) Not cars circulating only 5 days a week (which has sense if it happens to you to be a shopkeeper in a town built in the middle age waiting for deliveries)
USA is the place where everything is allowed, unless it's illegal.
USSR was the place where everything was forbidden, unless it was legal.
Italy is the place where everything is is allowed,
expecially what is forbidden.
AFAIK, Italy is the only nation where laws distinguish between "forbidden" and "severely forbidden, as if there were different grades of "forbidden". THIS is dumb.
My 0.01428 Euro this time (0.02 * 5/7).
Please note that this habit (of referring to X as the "X WindowS") had made it even in official GNU documentation (i.e. "X Windows" everywhere in the GNU Emacs 20.4 manual).
Perhaps it was done intentionally? Perhaps it's time to give it up, and add another row to the manpage?
At last, I see that using the right term is a way to show some respect to those people that provided us all with this great piece of well-engineered and well-thought software (...yes, I know that NeWS was technically better...).
Just a 0.01 Euro this time.
> Why not just support internationalization through gettext?
:-)
Because GNU gettext is under GPL, so...
But then, there is a more serious issue than that: actually, the framework they are using doesn't do any layout on dialogs, so it's likely that longer strings on labels/buttons/etc. would be displayed only partially.
Yes, it's definively a bad idea, but this is a trick to obtain more refresh speed, and AFAIK it's quite common in the Windows world, where geometry recalculation is done once, or none at all, and this is also the reason why UNIX toolkits seem slower. Actually, toolkit like GTK, or Motif, or Qt manage the layout by default (and the fixed-width/fixed position buttons are the exceptions, not the rule).
Last, StarOffice comes with a nice spellchecker, and that has to be made up for every single language (simply checking single words is definitively not enough in some languages like italian).
> False results. If the data format etc. are known, it's possible to feed the servers bogus results, which could lead to
inconsistencies in the data base
Send the same data to multiple receivers (randomly chosen), and see if they produce the same results. (or, at least, choerent ones). If note, one (or possibly more) are lying. Anyway, a closed-source client does not prevent someone to see what it does and send bogus data anyway. It only makes things harder for the ones that actually want to send correctd data.
> Data Theft. An open source program could be modified by Big Bad Corporation Inc. to simply harvest raw data and feed it
into their own computers
This is a more realistic issue, but Big Bad Corporation is probably rich enough to do reverse engineering of the protocol by itself, and access random lumps of raw data anyway. A closed-source client don't make much sense here.
The real point is that modified versions (i.e. to improve performance) could quickly spread so that just a few uses the original clients.
If suddenly it turns out that a widespread modified version produces erroneous data from time to time, then probably large amount of computations has to be thrown away. Of course, you could check for that using the same method you use to check for "bad guys", but it's a serious problem if you got only few people running the original.
My 0.02 Euro as usual.
> It was done in the /usr/bin/netscape script as a forced command line parameter when the browser is opened without any other
parameters
Well, SuSE also does this, but only if you didn't set up a homepage (that is, the wrapper checks for preferences.js and act accordingly).
But then, who really cares?
> GraphOn produces networking products, not Windows emulation. This is basically a Windows version of X-Windows
From what I understand reading on the GraphOn website, it seems to me that their product does exactly the job done by the well-known (and GPL'd) VNC, perhaps better.
> Windows by default uses this kind of approach, no need to use "extension"
FYI, at the time X11 was first released (1987), designers were broaded-mind enough to leave room for extensions of the protocol and of transport means. Today, SHM support is a de-facto standard on X servers (both commercial and XFree86), but it is still called 'extension' with respect to the protocol as defined in 1987.
Btw, I'd suggest you to not underestimate the capability of working over networks. That is exactly what many companies are wanting to sell as an innovation (see Windows Terminal server, Metaframe, PC/Anywhere, thin clients and so on), and X protocol is definitively faster (because only primitives has to be transfered, not bitmaps) on this side. Have a nice day.
> just disable all access to the account for 24 hours if they enter a bad password 3 times in 24 hours
Uhm, but then this may lead to DOS attacks, although this could be far less interesting for malicious people.
> MS is getting to be less and les a monopoly, because people are voting with their dollars
Although the scenery you prospect is really desiderable, please stop a moment to think about who has money to spend, and who has the power to decide where it should be spent.
In a perfect world, this wouldn't be a problem. In real world, there are people that have no choice but to accept what the market gives to them, because they may afford only the cheapest things. So, the cheapest wins.
Don't you think that a judgement such that was possibile because millions decided to give away their precious software for free with respect to freedom, breaking the market well-established rules?
My 0.0002 Euro, as usual.
> I'm not sure that you can
distribute kernel modules without some kind of GPL
As an exception Linus made, you may distribute binary-only modules, and the kernel has not to be modified in any way. But then the maintenance is completely up to you. This is how it has been done for the SBLive drivers, until Creative released the sources.
Please note that modules binaries are strongly kernel-version dependant, so if you provide a binary-only module, you'd provide it at least for more than one version of the kernel. This usually is bad (because you are stick with a few kernel versions supported), but IMHO it's not so bad in this case (a firewall machine is a firewall and nothing else, usually).
My 0.02 Euro.
> As I understand it, VNC is less like X and more of a constant-screengrab thing
Yes and no.
The server part on UNIX machines is just like another X server, only it pipes data through the net instead of video). Of course, you can have as many as you want running on a UNIX system (provided you have enough resources).
OTOH, on Windows and MacOS, it grabs the whole screen, so it may be useful only as a remote administering tool (ala Pc Anywhere).
The protocol being used actually transfers images in a way which seems to be quite efficient (but far from the efficiency obtained transfering mostly primitives). Still, it's really usable on a LAN.
There are platform-specific clients, and there's an implementation as a Java applet that just runs fine in a web browser. The interesting thing is that the VNC server part provides also an http interface that lets a user with only a web browser to authenticate and download&execute the Java client on the fly.
Actually, have you tried Sketch?
It runs pretty smoothly, and it's quite powerful for everyday's needs.
> I'm sure emacs could be programmed to be a full featured IDE
:-) it's already a full-featured IDE, expecially when you use the OO Browser from BeOpen. Surely, it takes some time to learn to use it efficiently. It does not slam on your face every feature it has, it may seem that it behaves strangely compared to other similar products, but don't forget that it comes out from over 20 years of experience and improvements.
:-)
I don't really want to start another holy war here, but for me (and ERS too
Oh, yeah, I don't mind if Linus considers it "pure evil"
> China might well become the last MS stronghold in the years to come :)
Hmmmm, I already see that: something like "A billion of users should be enough for anybody?"
> remember that WinAmp does a buttload more than x11amp
Maybe I am wrong, but have you recently visited the XMMS home?
It's a long time since x11Amp switched to XMMS.
> shareware like WinZip and the like might get ported... and I'm sure someone at WinAmp
would take a swat at it...
Bad examples: both exists already as native free application on Linux (well, actually not clones, but they provide the same user interface and the same functionalities). Probabily, there would be little or no commercial interests at all in porting applications like these, because the free ones are most of the times already there. Let's call it "natural software selection": the better ones are used at large, the worst ones simply die, or are used only by a restricted number of people that have a good reason to use them. Surely, not an interesting target for shareware makers (which usually need a large user base to be motivated, or at least the perspective of having a large paying user base, otherwise they'd used a free license).
But then, was the win32 API designed with portability in mind? Is there actually the will to make/keep it portable? I don't think so. That's seems to me just a temporary patch, not a real solution.
AFAIK, the current stable version of KDE cannot use the Qt 2.x libraries. The QPL licence covers only Qt releases >= 2.x.
Obviously, it doesn't make much sense to include KDE (which is free) if you can't use it.
> Stats I have seen (but have no pointers to) say CORBA is twice as popular as DCOM.
Surely, all telco companies use CORBA-based distributed systems. AFAIK, the rare times a DCOM interface is offered in these circumnstaces is just a 'plus'.
And then, there is already a standard protocol (IOOP) and a standard encapsulation in http (IOOP on HTTP).
So are we going to get also scrollbars that see in advance which part of those _damned_ fixed width dialog windows the user wants to see? Scrollbars are elevators, after all :-)
At least, it would be useful. But no, I wouldn't like scrollbars going on strike and feel depressed because they only go up and down (or left and right).
Premise: I use both Linux and HP_UX, and I like both of them. Evidently, Sanyo does not think this way.
Excerpts from the article above:
My 0.02 Euro, as usual