One problem is the following syllogism: a) a testing lab won't exist if it loses money b) a testing lab wil try to make money c) there will be 'competition' in the 'testing-labs' market d) You'll not know one 'works in linux' logo from the next.
I think it's probably better for it to remain in the hands of the distributions. Their names and their support structures are already in place.
Until then the best thing we can do is to a) write to mailing lists about incompatabilities found, keep everyone informed; then b) vote with your wallet. Don't buy those that lie or only pay lip-service to their support claims.
Take your choice indeed. I chose Finland 6 years ago, erroniously moved back to England, and recently returned. The Finns have been internet banking for a decade (and 10 years ago they were using higher security than Barclays Bank in England do now); there's an impressively high ratio of 'connected' poeple. Alternative (non x86/windows) systems are also popular (it's nothing to do with Linus' Finnishness, believe me). It's generally an IT-friendly place
Lots of good replies so far, yours included, most saying the similar things. I think one aspect has been missed though...
I for one haven't been as a consumer to a trade show since 1984 when I was heavily into the ZX Spectrum. I've been as an exhibitor since, so I've seen a fair few. My conclusions? (I hope I'm not alone.) I really don't think I get that much from trade shows. Not all 'geeks' need to go and see their 'heroes' speak, or to get some marketting low-down on the latest release BogoServ 1.9B. Attendence just ain't worth it (for me) any more. Fatigue's set in perhaps? Of course, geography, demographics and chronology didn't help this one either it seems.
Apple "lie with statistics". Using the data on the page you reference, I calculate the G4 to be 1.9 times as fast on average rather than 2.2 times. I notice what 'average' they use is not mentioned, as they know they're using one which is frowned upon.
If you take a 6 minute PIII job, consisting of equal time proportions of the 6 benchmark programs, and converted that to G4, it would only run 1.7 times as fast.
(Apple use arithmetic, spec (http://www.spec.org/osg/faq/) and people who know what they are doing use geometric, my final example uses harmonic)
I'm not saying 90% faster isn't a bad thing, BTW, simply narked by Apple's 'mathematics'.
Good point well made. Horses for courses, as we say.
A change in architecture doesn't necessarily mean times spent working with the old architecture is 'wasted'.
After developing some number-crunching code using GMP on a 21164 Alpha running Linux. I tried a piece of hand-coded assembly x86 Windows software which does the same job under Fx32! in NT.
The x86 code, _emulated_, was 3 times faster than what I could do natively. (533MHz 21164 runs like a 266-300Mhz PII)
Hats off to Yves Gallots optimisations, _and_ the Digital Fx32! developers.
I disagree. It is the compilers job to separate the programmer from the operating system. If the underlying allocator in the OS is abyssmal, the GCC is perfectly within its rights to shield the user from this by providing a superior heap. The user shouldn't even need to know whether this is happening or not. If gcc's malloc/realloc/free/new/delete is suboptimal because it relies directly on the OS, then it is only gcc to blame.
Phil
"Hey f*** off big-titted woman, we're talking about _software_" - Finnish man to Finnish lady in pub 2 nights back
If you ask me the guys who came up with repeated reports of cold fusion after it had started being debunked should be dragged through the streets and forced to repay every penny of taxpayers' money that they got in grants. OK, the guys who did it initially made a mistake, you can only prush boundaries if you step outside them. I'm thinking of the bloody bandwagon that followed.
FP. --
Real Men Don't Use Porn. -- Morality In Media Billboards
If he can say "not my fault, blame the TCP stack", then I agree with him. Even if he needs to change his code to get around weaknesses in other parts of the system I still don't think it's fair to call it a bug in his code.
FP
--
Real Men Don't Use Porn. -- Morality In Media Billboards
In the past I always blamed bad management, bad resourcing, bad marketing for the failures I have seen in the IT industries where I've worked.
In all these cases (I've worked in 10 places I guess now, that's a reasonable size) my blame was 100% placed in the right direction. (And I wasn't tactful in my 'leaving interviews' either - I once told HR that unless they fire Brian the manager there souldn't be a softy left on the project in 6 months, and they didn't believe me (fools)).
Finally I've found a place where it's the softies are to blame. Huge project. 10 softies, none of which could program in the language of the project. By heck did they right shit code. It's the softies' faults. But who was to blame for hiring them, not training them, and not managing them properly. Alas the upper levels have got to take the blame too.
In all these places, methodology was irrelevant. Common sense would have been relevant. None was to be found.
This really is a -1, Flamebait, subject, isn't it?
FP. --
Real Men Don't Use Porn. -- Morality In Media Billboards
Firstly take your bloody AC mask off. Are you embarassed by what you have said?
I fully understand the meaning of 'security through obscurity' as it is commonly used.
Notice how I deliberately _didn't_ use the phrase.
Yes, read my post, I do NOT use that phrase.
Then I suggest you look up 'obscure' in a dictionary -
"to conceal, to hide" or similar.
The data on freenet _is_ hidden. It is _concealed_. It is _obscured_. That's the point of it!
Notice also that I was only talking about the _distributed_ nature. The static model, and the communications protocols (the crypto) weren't being addressed at all.
I stand by what I said.
FatPhil --
Real Men Don't Use Porn. -- Morality In Media Billboards
Slashdot Mirror
One problem is the following syllogism:
a) a testing lab won't exist if it loses money
b) a testing lab wil try to make money
c) there will be 'competition' in the 'testing-labs' market
d) You'll not know one 'works in linux' logo from the next.
I think it's probably better for it to remain in the hands of the distributions. Their names and their support structures are already in place.
Until then the best thing we can do is to
a) write to mailing lists about incompatabilities found, keep everyone informed; then
b) vote with your wallet. Don't buy those that lie or only pay lip-service to their support claims.
FatPhil
Take your choice indeed.
I chose Finland 6 years ago, erroniously moved back to England, and recently returned.
The Finns have been internet banking for a decade (and 10 years ago they were using higher security than Barclays Bank in England do now); there's an impressively high ratio of 'connected' poeple. Alternative (non x86/windows) systems are also popular (it's nothing to do with Linus' Finnishness, believe me). It's generally an IT-friendly place
FatPhil
Lots of good replies so far, yours included, most saying the similar things. I think one aspect has been missed though...
I for one haven't been as a consumer to a trade show since 1984 when I was heavily into the ZX Spectrum. I've been as an exhibitor since, so I've seen a fair few.
My conclusions? (I hope I'm not alone.) I really don't think I get that much from trade shows. Not all 'geeks' need to go and see their 'heroes' speak, or to get some marketting low-down on the latest release BogoServ 1.9B. Attendence just ain't worth it (for me) any more. Fatigue's set in perhaps?
Of course, geography, demographics and chronology didn't help this one either it seems.
FatPhil
Apple "lie with statistics".
Using the data on the page you reference, I calculate the G4 to be 1.9 times as fast on average rather than 2.2 times. I notice what 'average' they use is not mentioned, as they know they're using one which is frowned upon.
If you take a 6 minute PIII job, consisting of equal time proportions of the 6 benchmark programs, and converted that to G4, it would only run 1.7 times as fast.
(Apple use arithmetic, spec (http://www.spec.org/osg/faq/) and people who know what they are doing use geometric, my final example uses harmonic)
I'm not saying 90% faster isn't a bad thing, BTW, simply narked by Apple's 'mathematics'.
FatPhil
(A Master of Arts in Mathematics, honest)
Good point well made. Horses for courses, as we say.
A change in architecture doesn't necessarily mean times spent working with the old architecture is 'wasted'.
After developing some number-crunching code using GMP on a 21164 Alpha running Linux. I tried a piece of hand-coded assembly x86 Windows software which does the same job under Fx32! in NT.
The x86 code, _emulated_, was 3 times faster than what I could do natively. (533MHz 21164 runs like a 266-300Mhz PII)
Hats off to Yves Gallots optimisations, _and_ the Digital Fx32! developers.
FatPhil
I'd bet on Intel. Two reasons:
1) Inertia
2) Momentum
Bugger, they're the same reason.
Last 3 processors bought by me?
K6-2
21164 (yup, I've been 64bit for years)
Athlon.
I'm voting with my wallet. The next machine? Dunno, but not Intel, I'm sure.
We're years away from a truly level playing field.
FatPhil
I disagree.
It is the compilers job to separate the programmer from the operating system. If the underlying allocator in the OS is abyssmal, the GCC is perfectly within its rights to shield the user from this by providing a superior heap. The user shouldn't even need to know whether this is happening or not. If gcc's malloc/realloc/free/new/delete is suboptimal because it relies directly on the OS, then it is only gcc to blame.
Phil
"Hey f*** off big-titted woman, we're talking about _software_"
- Finnish man to Finnish lady in pub 2 nights back
If you ask me the guys who came up with repeated reports of cold fusion after it had started being debunked should be dragged through the streets and forced to repay every penny of taxpayers' money that they got in grants. OK, the guys who did it initially made a mistake, you can only prush boundaries if you step outside them. I'm thinking of the bloody bandwagon that followed.
FP.
-- Real Men Don't Use Porn. -- Morality In Media Billboards
If he can say "not my fault, blame the TCP stack", then I agree with him. Even if he needs to change his code to get around weaknesses in other parts of the system I still don't think it's fair to call it a bug in his code.
FP
-- Real Men Don't Use Porn. -- Morality In Media Billboards
Blimey, I should spell-check, and proof-read.
s/souldn't/wouldn't/
s/right/write/
s/properly./properly?/
Sorry,
Phil
-- Real Men Don't Use Porn. -- Morality In Media Billboards
In the past I always blamed bad management, bad resourcing, bad marketing for the failures I have seen in the IT industries where I've worked.
In all these cases (I've worked in 10 places I guess now, that's a reasonable size) my blame was 100% placed in the right direction. (And I wasn't tactful in my 'leaving interviews' either - I once told HR that unless they fire Brian the manager there souldn't be a softy left on the project in 6 months, and they didn't believe me (fools)).
Finally I've found a place where it's the softies are to blame. Huge project. 10 softies, none of which could program in the language of the project. By heck did they right shit code. It's the softies' faults. But who was to blame for hiring them, not training them, and not managing them properly. Alas the upper levels have got to take the blame too.
In all these places, methodology was irrelevant. Common sense would have been relevant. None was to be found.
This really is a -1, Flamebait, subject, isn't it?
FP.
-- Real Men Don't Use Porn. -- Morality In Media Billboards
Firstly take your bloody AC mask off. Are you embarassed by what you have said?
I fully understand the meaning of 'security through obscurity' as it is commonly used.
Notice how I deliberately _didn't_ use the phrase.
Yes, read my post, I do NOT use that phrase.
Then I suggest you look up 'obscure' in a dictionary -
"to conceal, to hide" or similar.
The data on freenet _is_ hidden. It is _concealed_. It is _obscured_. That's the point of it!
Notice also that I was only talking about the _distributed_ nature. The static model, and the communications protocols (the crypto) weren't being addressed at all.
I stand by what I said.
FatPhil
-- Real Men Don't Use Porn. -- Morality In Media Billboards