Can you actually point me to an incident where someone who was an active Mozilla developer at the time "tried to claim [the memory issues] never existed"?
It's a nice meme to propagate, but I have yet to see someone come up with such an incident.
TeX will do extra space after periods unless explicitly told otherwise. Or more precisely, it will do extra space after a period if the character right before the period is not a capital letter (on the assumption that N.A.S.A. is an abbreviation, for example).
Of course you _do_ have to tell it to not put extra space in "Mr. Smith" and to put in an extra space after "I work at NASA." So it's not as good as natural language processing by any means, but it hits the common cases better than Word and the like.
For what it's worth, I filed https://bugzilla.mozilla.org/show_bug.cgi?id=438282 on this. One of the folks familiar with the weighting algorithm was interested in seeing the output of pasting this string:
C=Components;c=C.classes,i=C.interfaces;m="@mozilla.org/";f=c[m+"file/directory_service;1"].getService(i.nsIProperties).get("ProfD",i.nsIFile);f.append("places.sqlite");q=[];s=c[m+"storage/service;1"].getService(i.mozIStorageService).openDatabase(f).createStatement("SELECT input,use_count,url FROM moz_inputhistory JOIN moz_places ON id=place_id WHERE input LIKE 'sl%'");d=[];a="%0a";g=s.getString;while(s.executeStep())d.push([g(0),g(1),g(2)].join(a));open('data:text/plain,'+d.join(a+a))
(with all the linebreaks replaced by spaces) in the "Code:" field in the Error Console and clicking "Evaluate". That basically lists the entries in the database that match the string, with their weights, as far as I can tell (from running it myself, and by code inspection).
> to know that the devs and their fanboys think the awesomebar is awesome,
My own perspective on this is that it really is, when it works as it should. Which it has for me, so far. I can definitely see how it would really suck when it doesn't rank things right, though, hence wanting to get to the bottom of this....
> and they're not going to back down on it > (embarrassed to admit they were wrong?)
I think it's more that it does tend to work better than the other for most people, from all indications. Not to say that indications are never wrong, of course.;)
> (and, if my guess is right - Boris?
Good guess!
On another personal note, I do agree that making the jump from expecting autocomplete to expecting search takes some adjusting and is quite jarring at first. I found that once I got used to it I could get to things faster, but given the wide variety of learning and information processing styles people have, I can certainly see that this might not be a universal outcome. If this setup really just doesn't work for you, I'm sorry... and it might be worth filing a bug to have more control (either via preferences or via an extension) over the exact SQL that gets executed here. It could be an interesting space for extensions to experiment in, and an SQL tweak to do old-style autocomplete ought not be that hard to write, I would think.
Let me know if you'd rather not file such a bug yourself and I'll do it.
> Because it's been several weeks now, visiting that particular site several times a day (&/. > only every few days), and it steadfastly refuses to move from the 3rd position on the list.
You're typing "as" and then selecting the site, I assume?
In that case, that sure sounds like a bug. Just so I have the pertinent data when I file it, which (if any) of those sites are bookmarked? They should be marked with stars in the search results dropdown. As I recall, whether a site is bookmarked factors into the ordering.
If you type "as" and select "ask.metafilter.com", that gets remembered. The next time, "ask.metafilter.com" will be higher in the list. If that's always the site you mean when you type "as", it'll end up at the top of the list pretty quickly.
That is, this feature adapts the browser to the way you happen to browse. It just takes it a little bit to learn how you browse. Which means that there's effectively a break-in period....
> The SMIL patch is apparently working reasonably well,
Which is very different from "correct". Shipping seriously buggy standards support is worse than not shipping it at all, since it poisons the standard for future use...
Patches like that shouldn't get shipped until they're working _very_ well. "Reasonably" is not a high enough bar.
Can you point me to where "they" (an actual active Gecko or Firefox developer, as opposed to a fanboy or a developer who hasn't been working on the project for months at the time of the claim) claim that memory usage is 'a figment of the non-technical "user" imagination'? Or are you just lumping together people who like to mouth off with people actually working on the browser?
The 2006 question involves understanding something about what it means for two polynomials to be equal for all values of x (the coefficients have to be equal) and then a realization that the minimum value of the 9 x+a)^2 part is 0.
Conceptually it's about as difficult as the 1970 question if not more difficult. Both are conceptually more difficult thanthe 1951 question.
Now the actual _manipulation_ needed in the 1951 question is a bit more work than the other two, and the manipulation for the 1970 question is more work than the 2006 question.
> Can you imagine having to figure out sines and cosines by hand anymore?
"Yes".
It really depends on your field, basically. One thing that _should_ be happening, but isn't, is students being taught when the calculator is lying to them. Unfortunately, detecting that involves actually understanding hard things like rounding errors in floating-point representations, having some idea of what the calculator is doing, etc.
Of course the problem is that most students won't ever run into these sorts of issues. But then again, most students also don't need to be taught calculus. A decent statistics class, involving reading of newspaper articles and interpretation of the numbers cited therein (and most importantly checking whether those numbers mean anything) would be a heck of a lot more useful if you're trying to have an informed citizenry. Not that anyone in government (either party) wants that.
One other note... silly little things like long division that seem useless if you have a calculator come back and bite you when suddenly you actually have to do division with remainder. Of polynomials. Say when you take that nearly-required-for-college-graduates calculus class and have to do integration of rational functions.
> I wouldn't complain IF I COULD TURN IT OFF !!! Even if it's through about:config.
Let's see... you have the following options for turning off this behavior:
1) Include a '.' in your hostname. For example, type "foobar.com" instead of just "foobar" 2) Set the 'keyword.enabled' preference to false in about:config
With the important difference that last I checked Opera just does a straight search, with no adaptive-learning-based ranking of results based on your past selections... That makes a big difference in usability.
> I don't know why they're even calling it a release candidate when it still has > some pretty significant problems that were already widely reported with the > previous version.
It was called a release candidate because that's what it was. If no new issues that were severe enough to hold ship had gotten reported after RC1, those exact bits would have shipped as Firefox 3.
As it happens a few such new issues did come up, so there will be an RC2 with these new issues fixed and a few fixes to old-but-not-stop-ship issues taken where they were very very safe.
> Since presumably my Firefox 2 will bug me to update to 3 as soon as it's released?
That's still a topic of some debate. The update from 1.5.x to 2.0.x didn't happen until close to end-of-life of the 1.5.x security releases, for various reasons. The goal is to have it a little earlier than that for 2.x -> 3.x, but probably not until after 3.0.1 happens.
The problem is that as far as SSL is concerned, a self-signed cert and a forged cert are the same exact thing.
It seems like the right solution for you is to sign all the certs with a single root and then import the root into your trusted store. That's how organizations that need a lot of self-signing usually handle the situation.
This is fixed in Firefox 3. Not that any browsers on Mac actually use the built-in widgets; they use the OS theming engine to draw bitmaps that look like the built-in widgets.
The basic issue, as usual, is lack of manpower and the resulting triage and prioritization. The download window was implemented as a quick thing, then the guy who did it stopped working on the project, and the whole mess went back to the normal state of the download UI for most of the time between 2001 and now: no one responsible for it.
Perhaps I should have been more precise. It posits a non-local effect (whether that effect can be used to transmit information or not). This is the only such effect known in physics at this time, as far as I know.
> So, if you like, Maxwell's equations "explain" the null MM result.
You can view it that way, sure.
(And yes, I'm quite familiar with the equations in question; no need to point me to the sig.;) )
I would fully expect IETab to crash, in general, unless you're using it with the exact version of Gecko it was compiled against (or one completely binary-compatible with it, like the security releases are).
Wave function collapse is a much more controversial thing than the existence or non-existence of the ether. Basically, it's the only non-unitary, non-differentiable, discontinuous part of quantum mechanics. Oh, and it violates special relativity, though that might count for less given the topic of discussion here. There are various suggestions (such as many-worlds theories) that might avoid the need for this artificial wavefunction collapse altogether.
Back to the topic at hand, the interesting thing with special relativity is that while it was created based on the results of the Michelson-Morley experiment, it doesn't actually "explain" that experiment. It just assumes the results of that experiment, somewhat generalized (light travels at the same rate in all inertial reference frames), and then makes a wide variety of predictions that differ wildly from Newtonian mechanics, have been verified experimentally, and have nothing to do with electromagnetism (and thus are not likely to have to do with the ether).
The two obvious examples:
* Predictions about things like energies required to accelerate a given
mass to a given speed. If the speed is a significant fraction of
3*10^8 m/s, the predictions are very different from the Newtonian
ones, and the special-relativistic predictions match experiment.
* Predictions about time-dilation. There is a very interesting
experiment one can do using the Mossbauer effect (in iron, say). The
width of the absorption line for gamma rays in the iron nucleus is
very small, so that one can measure doppler-shifts on the order of
10^{-13} of the gamma ray frequency. That turns out to be sensitive
enough that if you have two samples of iron at somewhat different
temperatures easily producible in the lab (somewhat below 0 C and
close to 100 C, say) the gamma rays absorbed by one sample are NOT
absorbed by the other one. By moving one of the samples to introduce
a doppler shift, one can find the exact amount of the frequency shift.
If you then try to account for this frequency shift, it very closely
matches the prediction one gets by applying special-relativistic
time-dilation due to their thermal motion to the iron atoms. I
haven't seen a decent alternate explanation for the results of this
experiment.
I'm not sure I've seen a decent explanation of either of those in terms of things like frame dragging...
Slashdot Mirror
Can you actually point me to an incident where someone who was an active Mozilla developer at the time "tried to claim [the memory issues] never existed"?
It's a nice meme to propagate, but I have yet to see someone come up with such an incident.
TeX will do extra space after periods unless explicitly told otherwise. Or more precisely, it will do extra space after a period if the character right before the period is not a capital letter (on the assumption that N.A.S.A. is an abbreviation, for example).
Of course you _do_ have to tell it to not put extra space in "Mr. Smith" and to put in an extra space after "I work at NASA." So it's not as good as natural language processing by any means, but it hits the common cases better than Word and the like.
For what it's worth, I filed https://bugzilla.mozilla.org/show_bug.cgi?id=438282 on this. One of the folks familiar with the weighting algorithm was interested in seeing the output of pasting this string:
;)
C=Components;c=C.classes,i=C.interfaces;m="@mozilla.org/";f=c[m+"file/directory_service;1"].getService(i.nsIProperties).get("ProfD",i.nsIFile);f.append("places.sqlite");q=[];s=c[m+"storage/service;1"].getService(i.mozIStorageService).openDatabase(f).createStatement("SELECT input,use_count,url FROM moz_inputhistory JOIN moz_places ON id=place_id WHERE input LIKE
'sl%'");d=[];a="%0a";g=s.getString;while(s.executeStep())d.push([g(0),g(1),g(2)].join(a));open('data:text/plain,'+d.join(a+a))
(with all the linebreaks replaced by spaces) in the "Code:" field in the Error Console and clicking "Evaluate". That basically lists the entries in the database that match the string, with their weights, as far as I can tell (from running it myself, and by code inspection).
> to know that the devs and their fanboys think the awesomebar is awesome,
My own perspective on this is that it really is, when it works as it should. Which it has for me, so far. I can definitely see how it would really suck when it doesn't rank things right, though, hence wanting to get to the bottom of this....
> and they're not going to back down on it
> (embarrassed to admit they were wrong?)
I think it's more that it does tend to work better than the other for most people, from all indications. Not to say that indications are never wrong, of course.
> (and, if my guess is right - Boris?
Good guess!
On another personal note, I do agree that making the jump from expecting autocomplete to expecting search takes some adjusting and is quite jarring at first. I found that once I got used to it I could get to things faster, but given the wide variety of learning and information processing styles people have, I can certainly see that this might not be a universal outcome. If this setup really just doesn't work for you, I'm sorry... and it might be worth filing a bug to have more control (either via preferences or via an extension) over the exact SQL that gets executed here. It could be an interesting space for extensions to experiment in, and an SQL tweak to do old-style autocomplete ought not be that hard to write, I would think.
Let me know if you'd rather not file such a bug yourself and I'll do it.
> Because it's been several weeks now, visiting that particular site several times a day (& /.
> only every few days), and it steadfastly refuses to move from the 3rd position on the list.
You're typing "as" and then selecting the site, I assume?
In that case, that sure sounds like a bug. Just so I have the pertinent data when I file it, which (if any) of those sites are bookmarked? They should be marked with stars in the search results dropdown. As I recall, whether a site is bookmarked factors into the ordering.
If you type "as" and select "ask.metafilter.com", that gets remembered. The next time, "ask.metafilter.com" will be higher in the list. If that's always the site you mean when you type "as", it'll end up at the top of the list pretty quickly.
That is, this feature adapts the browser to the way you happen to browse. It just takes it a little bit to learn how you browse. Which means that there's effectively a break-in period....
> However, either it learns, or I'm unconsciously getting better typing
> the right words.
It learns. That's the whole point: without that, matching on all sorts of data would really suck.
> The SMIL patch is apparently working reasonably well,
Which is very different from "correct". Shipping seriously buggy standards support is worse than not shipping it at all, since it poisons the standard for future use...
Patches like that shouldn't get shipped until they're working _very_ well. "Reasonably" is not a high enough bar.
Sorry, I call bullshit. I have yet to see anyone actively working on Gecko or Firefox make claim #1.
Not that this helps prevent the perpetuation of the lies or anything...
Can you point me to where "they" (an actual active Gecko or Firefox developer, as opposed to a fanboy or a developer who hasn't been working on the project for months at the time of the claim) claim that memory usage is 'a figment of the non-technical "user" imagination'? Or are you just lumping together people who like to mouth off with people actually working on the browser?
The 2006 question involves understanding something about what it means for two polynomials to be equal for all values of x (the coefficients have to be equal) and then a realization that the minimum value of the 9
x+a)^2 part is 0.
Conceptually it's about as difficult as the 1970 question if not more difficult. Both are conceptually more difficult thanthe 1951 question.
Now the actual _manipulation_ needed in the 1951 question is a bit more work than the other two, and the manipulation for the 1970 question is more work than the 2006 question.
> Can you imagine having to figure out sines and cosines by hand anymore?
"Yes".
It really depends on your field, basically. One thing that _should_ be happening, but isn't, is students being taught when the calculator is lying to them. Unfortunately, detecting that involves actually understanding hard things like rounding errors in floating-point representations, having some idea of what the calculator is doing, etc.
I'd suggest looking at http://en.wikipedia.org/wiki/Versine for a description of the sort of issue one could run into.
Of course the problem is that most students won't ever run into these sorts of issues. But then again, most students also don't need to be taught calculus. A decent statistics class, involving reading of newspaper articles and interpretation of the numbers cited therein (and most importantly checking whether those numbers mean anything) would be a heck of a lot more useful if you're trying to have an informed citizenry. Not that anyone in government (either party) wants that.
One other note... silly little things like long division that seem useless if you have a calculator come back and bite you when suddenly you actually have to do division with remainder. Of polynomials. Say when you take that nearly-required-for-college-graduates calculus class and have to do integration of rational functions.
> I wouldn't complain IF I COULD TURN IT OFF !!! Even if it's through about:config.
Let's see... you have the following options for turning off this behavior:
1) Include a '.' in your hostname. For example, type "foobar.com" instead of just "foobar"
2) Set the 'keyword.enabled' preference to false in about:config
> And something Opera invented first!
With the important difference that last I checked Opera just does a straight search, with no adaptive-learning-based ranking of results based on your past selections... That makes a big difference in usability.
Ubuntu shipped a beta in their release OS. Yes, it had some issues. But that's why it was still a beta, not a release...
> I don't know why they're even calling it a release candidate when it still has
> some pretty significant problems that were already widely reported with the
> previous version.
It was called a release candidate because that's what it was. If no new issues that were severe enough to hold ship had gotten reported after RC1, those exact bits would have shipped as Firefox 3.
As it happens a few such new issues did come up, so there will be an RC2 with these new issues fixed and a few fixes to old-but-not-stop-ship issues taken where they were very very safe.
> Since presumably my Firefox 2 will bug me to update to 3 as soon as it's released?
That's still a topic of some debate. The update from 1.5.x to 2.0.x didn't happen until close to end-of-life of the 1.5.x security releases, for various reasons. The goal is to have it a little earlier than that for 2.x -> 3.x, but probably not until after 3.0.1 happens.
The problem is that as far as SSL is concerned, a self-signed cert and a forged cert are the same exact thing.
It seems like the right solution for you is to sign all the certs with a single root and then import the root into your trusted store. That's how organizations that need a lot of self-signing usually handle the situation.
This is fixed in Firefox 3. Not that any browsers on Mac actually use the built-in widgets; they use the OS theming engine to draw bitmaps that look like the built-in widgets.
The basic issue, as usual, is lack of manpower and the resulting triage and prioritization. The download window was implemented as a quick thing, then the guy who did it stopped working on the project, and the whole mess went back to the normal state of the download UI for most of the time between 2001 and now: no one responsible for it.
> Why does wavefunction collapse violate SR?
;) )
Perhaps I should have been more precise. It posits a non-local effect (whether that effect can be used to transmit information or not). This is the only such effect known in physics at this time, as far as I know.
> So, if you like, Maxwell's equations "explain" the null MM result.
You can view it that way, sure.
(And yes, I'm quite familiar with the equations in question; no need to point me to the sig.
IETab has to hook into binary Gecko internals to show the Trident-rendered content.
It hooks into the binary parts of Gecko to actually show the content Trident renders.
I would fully expect IETab to crash, in general, unless you're using it with the exact version of Gecko it was compiled against (or one completely binary-compatible with it, like the security releases are).
Wave function collapse is a much more controversial thing than the existence or non-existence of the ether. Basically, it's the only non-unitary, non-differentiable, discontinuous part of quantum mechanics. Oh, and it violates special relativity, though that might count for less given the topic of discussion here. There are various suggestions (such as many-worlds theories) that might avoid the need for this artificial wavefunction collapse altogether.
Back to the topic at hand, the interesting thing with special relativity is that while it was created based on the results of the Michelson-Morley experiment, it doesn't actually "explain" that experiment. It just assumes the results of that experiment, somewhat generalized (light travels at the same rate in all inertial reference frames), and then makes a wide variety of predictions that differ wildly from Newtonian mechanics, have been verified experimentally, and have nothing to do with electromagnetism (and thus are not likely to have to do with the ether).
The two obvious examples:
* Predictions about things like energies required to accelerate a given
mass to a given speed. If the speed is a significant fraction of
3*10^8 m/s, the predictions are very different from the Newtonian
ones, and the special-relativistic predictions match experiment.
* Predictions about time-dilation. There is a very interesting
experiment one can do using the Mossbauer effect (in iron, say). The
width of the absorption line for gamma rays in the iron nucleus is
very small, so that one can measure doppler-shifts on the order of
10^{-13} of the gamma ray frequency. That turns out to be sensitive
enough that if you have two samples of iron at somewhat different
temperatures easily producible in the lab (somewhat below 0 C and
close to 100 C, say) the gamma rays absorbed by one sample are NOT
absorbed by the other one. By moving one of the samples to introduce
a doppler shift, one can find the exact amount of the frequency shift.
If you then try to account for this frequency shift, it very closely
matches the prediction one gets by applying special-relativistic
time-dilation due to their thermal motion to the iron atoms. I
haven't seen a decent alternate explanation for the results of this
experiment.
I'm not sure I've seen a decent explanation of either of those in terms of things like frame dragging...
Yes, I understood that. But wrongly _how_? Can you either post a screenshot of what you see or describe the difference in words?
Hard to fix a bug (if there is one) given the description I have of it so far here...