Reply by dbd March 27, 20082008-03-27
On Mar 26, 10:52 pm, robert bristow-johnson
<r...@audioimagination.com> wrote:
> On Mar 26, 9:11 pm, dbd <d...@ieee.org> wrote:
> ... > > I'll give you "n+0 = n", but point by point discussion of points we > > don't even agree on the definitions of won't help. It won't deal with > > the issues of completeness, compatibility and side effects the way an > > instantiation will. > > the issue of completeness is valid BUT that is NOT the same as the > issue of backward compatibility. > ... > now, completeness, that's a different story, but we would have time to > develop meaningful extensions to existing operations on arrays that > have origins other than 1. in my treatise posted on comp.soft- > sys.matlab 8 years ago, i described how this would work specifically > for array (or matrix) addition, subtraction, even matrix > multiplication and the backslash operator (at least for square > matrices). also for array concantenation, and for functions that > return indices like min(), max(), and find(). that is all well > defined in a meaningful way that defaults to the existing definitions > when the origin of each dimension is 1. > > so then, WHAT'S THE PROBLEM??? why such resistance? i think, for > MathWorks, it's a virulent case of the "NIH syndrome", which i have to > confess exists here at Kurzweil Music Systems also, much to its > dettriment. maybe it's a Boston thing, is NIH ("Not Invented Here") a > problem for you guys out at the left coast? or you folks in Canada or > Europe or the East? it's really stupid and counter-productive. >
I still think your description is incomplete and broken, but I'd be happy to let you fix that and "invent it here". If you can I'd love to test it. If it works I'd use it. Don't you have any faith in the workability of your conception? WHAT'S THE PROBLEM??? (see you can capitalize) Why the resistance?
> > There are a lot of modelers here. You think it is > > simple. > > ... > > Build the model. We'll play with it. If it works, it will be a > > much better lever to move Mathworks with than your declaration of > > being boggled in 2008 that you haven't gotten your wish yet. > > as i said, I CAN'T DO IT because the Octave source is (shamefully) a > spaghetti bowl (designed by committee, the GNU way, and if anyone > complains, tell them "you get what you pay for" - when my PC here at > work got converted to Redhat, the DVD player was permanently rendered > useless - take a look athttp://www.youtube.com/watch?v=h4iyksLeo7w > for some laffs) and i don't have access to the MATLAB source, of > course. > > r b-j
I've long since given up suggesting you work with the real code. I suggested a shell around Matlab that would only need to do the re- indexing. You don't need to see or understand the proprietary hooks to do that. Build it around Octave, if you choose. Pick a convincingly useful set of operations to support. You can pick your choice of language and platform for the shell, too. You Can Invent It Here! (YCIIH) That's acceptance not resistance. You say you want it, it's important, it's simple and you understand how it works. If you don't care enough to back that up why should anyone consider your arguments credible? The Octave code may have been "designed by a committee", but the committee isn't here posting messages saying "Why hasn't someone written this for me?", waiting for Octave to appear. Dale B. Dalrymple
Reply by robert bristow-johnson March 27, 20082008-03-27
On Mar 26, 9:11 pm, dbd <d...@ieee.org> wrote:
> On Mar 25, 10:27 pm, robert bristow-johnson > > <r...@audioimagination.com> wrote: > >.. > > > did you actually read the proposal, Dale? you said you did. > > Yes I had read it. Now I've even read it again. Now I think it is > broken and has errors. > > I still don't believe that your 'constraints' are consistent with what > I consider to be backward compatibility. > > > ... > > the whiners are the people saying "oh me, oh my, it will break all of > > our existing code" to which i say, "no, it won't, just as adding 0 to > > a number does not change it." > > I thought whiners were people who said "I don't know where to begin" > but then do anyway:) > > > > > Dale, if you're willing to, i'll debate this technically point-by- > > point. i am saying that for most basic function, this extension is > > well defined, and will break NO existing MATLAB (or Octave) source > > code, because adding zero doesn't change anything. > > > r b-j > > I'll give you "n+0 = n", but point by point discussion of points we > don't even agree on the definitions of won't help. It won't deal with > the issues of completeness, compatibility and side effects the way an > instantiation will.
the issue of completeness is valid BUT that is NOT the same as the issue of backward compatibility. my claim 8 years ago was that this fix or enhancement can be made, reasonably painlessly, to the MATLAB (or Octave) executable by someone who is familiar with the source, in such a way to GUARANTEE backward compatibility. then no one can bitch because it broke their existing MATLAB (or Octave) program. that should be sufficient to sweep aside objections for that reason, but, amazingly it did not (and i have little respect for those objections, because they insult my intelligence). now, completeness, that's a different story, but we would have time to develop meaningful extensions to existing operations on arrays that have origins other than 1. in my treatise posted on comp.soft- sys.matlab 8 years ago, i described how this would work specifically for array (or matrix) addition, subtraction, even matrix multiplication and the backslash operator (at least for square matrices). also for array concantenation, and for functions that return indices like min(), max(), and find(). that is all well defined in a meaningful way that defaults to the existing definitions when the origin of each dimension is 1. so then, WHAT'S THE PROBLEM??? why such resistance? i think, for MathWorks, it's a virulent case of the "NIH syndrome", which i have to confess exists here at Kurzweil Music Systems also, much to its dettriment. maybe it's a Boston thing, is NIH ("Not Invented Here") a problem for you guys out at the left coast? or you folks in Canada or Europe or the East? it's really stupid and counter-productive.
> There are a lot of modelers here. You think it is > simple.
i think the basic concept is simple, implementation of it for someone who knows where all the indexing hooks are, is not a huge project, and that backward compatibility can be guaranteed. defining it for the most basic operations is reasonably straight-forward and extending it to other functions (the completeness issue) can be developed in time.
> Build the model. We'll play with it. If it works, it will be a > much better lever to move Mathworks with than your declaration of > being boggled in 2008 that you haven't gotten your wish yet.
as i said, I CAN'T DO IT because the Octave source is (shamefully) a spaghetti bowl (designed by committee, the GNU way, and if anyone complains, tell them "you get what you pay for" - when my PC here at work got converted to Redhat, the DVD player was permanently rendered useless - take a look at http://www.youtube.com/watch?v=h4iyksLeo7w for some laffs) and i don't have access to the MATLAB source, of course. r b-j
Reply by dbd March 26, 20082008-03-26
On Mar 25, 10:27 pm, robert bristow-johnson
<r...@audioimagination.com> wrote:

>.. > > did you actually read the proposal, Dale? you said you did. >
Yes I had read it. Now I've even read it again. Now I think it is broken and has errors. I still don't believe that your 'constraints' are consistent with what I consider to be backward compatibility.
> ... > the whiners are the people saying "oh me, oh my, it will break all of > our existing code" to which i say, "no, it won't, just as adding 0 to > a number does not change it."
I thought whiners were people who said "I don't know where to begin" but then do anyway:)
> > Dale, if you're willing to, i'll debate this technically point-by- > point. i am saying that for most basic function, this extension is > well defined, and will break NO existing MATLAB (or Octave) source > code, because adding zero doesn't change anything. > > r b-j
I'll give you "n+0 = n", but point by point discussion of points we don't even agree on the definitions of won't help. It won't deal with the issues of completeness, compatibility and side effects the way an instantiation will. There are a lot of modelers here. You think it is simple. Build the model. We'll play with it. If it works, it will be a much better lever to move Mathworks with than your declaration of being boggled in 2008 that you haven't gotten your wish yet. Dale B. Dalrymple
Reply by Jerry Avins March 26, 20082008-03-26
Randy Yates wrote:
> Grant Griffin <nobody@example.com> writes: >> [...] >> Randy Yates wrote: >>> Again, you've narrowed in on an embarrassing lack in linux. It took >>> me 3 or 4 DAYs to get the Netgear wlan card working on my wife's Fedora >>> 8 system. I'm stuck with ndiswrapper, built custom for the kernel, along >>> with the XP driver files for the card, but it works. >> Kindda makes $100 for a copy of Windows seem cheap, don't it? ;-) > > Last time I checked, they were charging around $200 (or more) for > Vista. Oh yeah, and then you need to shuck out another $200-300 for > MSOffice, too.
I think Open Office is just fine. Better, in some respects.
> It really doesn't matter - they could give it away and I still wouldn't > use it. As costly as my experience was, there are a host of other issues > associated with a Redmond OS that would pester me every day of my life, > making it much more costly in the long run.
Jerry -- Engineering is the art of making what you want from things you can get. &#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;
Reply by Randy Yates March 26, 20082008-03-26
Grant Griffin <nobody@example.com> writes:
> [...] > Randy Yates wrote: >> Again, you've narrowed in on an embarrassing lack in linux. It took >> me 3 or 4 DAYs to get the Netgear wlan card working on my wife's Fedora >> 8 system. I'm stuck with ndiswrapper, built custom for the kernel, along >> with the XP driver files for the card, but it works. > > Kindda makes $100 for a copy of Windows seem cheap, don't it? ;-)
Last time I checked, they were charging around $200 (or more) for Vista. Oh yeah, and then you need to shuck out another $200-300 for MSOffice, too. It really doesn't matter - they could give it away and I still wouldn't use it. As costly as my experience was, there are a host of other issues associated with a Redmond OS that would pester me every day of my life, making it much more costly in the long run. -- % Randy Yates % "Maybe one day I'll feel her cold embrace, %% Fuquay-Varina, NC % and kiss her interface, %%% 919-577-9882 % til then, I'll leave her alone." %%%% <yates@ieee.org> % 'Yours Truly, 2095', *Time*, ELO http://www.digitalsignallabs.com
Reply by Scott Seidman March 26, 20082008-03-26
robert bristow-johnson <rbj@audioimagination.com> wrote in news:6fbca402-
9bea-4d1a-8fde-8aa09d79fd99@m36g2000hse.googlegroups.com:

> they changed the definition a *little* if X[1] is DC.
X[1] is the first element of the data stored in X, which corresponds to the DC value. The indexing scheme does not change the definition, and the definition does not change the indexing scheme. It can become a problem if you insist on assigning some sort of relationship between the index and the equation, but I haven't. It certainly can be a bit of a problem when going back and forth between languages, but I generally prototype in one then move to the next, and handle the conversions all at once. To extend the example, if X were two dimensional instead of one, since there's no pointer in Matlab, I couldn't care less (and in fact, I don't really recall) whether the arrays are row-indexed or column indexed, so long as I know what I'm going to get back if I call for X[i,j]. -- Scott Reverse name to reply
Reply by Scott Seidman March 26, 20082008-03-26
robert bristow-johnson <rbj@audioimagination.com> wrote in news:6fbca402-
9bea-4d1a-8fde-8aa09d79fd99@m36g2000hse.googlegroups.com:

> they changed the definition a *little* if X[1] is DC.
X[1] is the first element of the data stored in X, which corresponds to the DC value. The indexing scheme does not change the definition, and the definition does not change the indexing scheme. It can become a problem if you insist on assigning some sort of relationship between the index and the equation, but I haven't. It certainly can be a bit of a problem when going back and forth between languages, but I generally prototype in one then move to the next, and handle the conversions all at once. To extend the example, if X were two dimensional instead of one, since there's no pointer in Matlab, I couldn't care less (and in fact, I don't really recall) whether the arrays are row-indexed or column indexed, so long as I know what I'm going to get back if I call for X[i,j]. -- Scott Reverse name to reply
Reply by Scott Seidman March 26, 20082008-03-26
robert bristow-johnson <rbj@audioimagination.com> wrote in news:6fbca402-
9bea-4d1a-8fde-8aa09d79fd99@m36g2000hse.googlegroups.com:

> they changed the definition a *little* if X[1] is DC.
X[1] is the first element of the data stored in X, which corresponds to the DC value. The indexing scheme does not change the definition, and the definition does not change the indexing scheme. It can become a problem if you insist on assigning some sort of relationship between the index and the equation, but I haven't. It certainly can be a bit of a problem when going back and forth between languages, but I generally prototype in one then move to the next, and handle the conversions all at once. To extend the example, if X were two dimensional instead of one, since there's no pointer in Matlab, I couldn't care less (and in fact, I don't really recall) whether the arrays are row-indexed or column indexed, so long as I know what I'm going to get back if I call for X[i,j]. -- Scott Reverse name to reply
Reply by robert bristow-johnson March 26, 20082008-03-26
On Mar 26, 10:08&#2013266080;am, Scott Seidman <namdiestt...@mindspring.com>
wrote:
> robert bristow-johnson <r...@audioimagination.com> wrote in news:9bad46f9- > c54d-46a1-b67d-956cae2dd...@e39g2000hsf.googlegroups.com: > > > i've done it dozens of times in C code. &#2013266080;it all depends on where you > > say the "0th" element is. > > Am I the odd one here? &#2013266080;I've never seen the array index as anything more > than a mapping into some memory space, and never assigned it any magical > value in any equation. &#2013266080;Different languages access this memory a bit > differently, and sometimes it causes a little screwup when I lose track, > but I just don't see that a different indexing scheme changes the > fundamental definition of the FFT or anything like that.
they changed the definition a *little* if X[1] is DC. Scott, it's a usage thing. if MATLAB lived up to their self-imposed reputation (allowing us to write equations in MATLAB that appear like the equations we write on paper), then we shouldn't always have to subtract 1 from the subscript returned from max() or min() or find(). we should not always have to add 1 to the subscript when doing synthesis just before calling the IFFT. it's stupid. what i fear is that textbooks will start changing their definitions of the DFT or discrete convolution to accomodate MATLAB. that would be *really* stupid. r b-j
Reply by dbd March 26, 20082008-03-26
On Mar 26, 7:08 am, Scott Seidman <namdiestt...@mindspring.com> wrote:
> ... > Different languages access this memory a bit > differently, and sometimes it causes a little screwup when I lose track, > but I just don't see that a different indexing scheme changes the > fundamental definition of the FFT or anything like that. > -- > Scott
Because I have written more code in C than in Fortran, I think 0 based indexing in Matlab would have been convenient. I don't think it is a necessity. The number of Matlab users proves that it isn't. From earlier in this discussion: & (to me, it's just astonishing that it's 2008 and we don't &have an fft() in the most common signal processing model/emulating &software that returns the DC value in bin #0! that's just stupid, and &like some other software shit, compared to the 1990s, i think our &tools today are ickyer and harder to use than in the 90s. i guess i'm &just a Mac wus, but even the OSX Mac stuff is harder to deal with than &what we were using in the olden days.) & &r b-j I think generalized indexing would be the kind of thing that would make Matlab "ickyer and harder to use". I think it would result in code that expresses the same thing in many different ways which would make code harder to reuse. It would more the semantic arguments from comp.dsp to comp.soft-sys.matlab... Hey! Maybe it is a good idea! But if it's as simple as some people say, I'm willing to try an test implementation to see if I'm wrong and nothing breaks, etc. And in the 1990s didn't we use capitalization for readability? But in the spirit of religious discussions, which this topic has long since become... From: http://groups.google.com/group/comp.soft-sys.matlab/msg/1189e49bb728a4fd
>It boggles my mind that a >developer apparently refuses to recognize that they have the potential >to increase the scope of their tool to an obviously large class of >potential customers. > >>Yes, many of us at MathWorks continue to follow this thread. > >Unfortunately, it seems that is all that is ever done. > >Eric Jacobsen
Some of God's greatest gifts are unanswered prayers Garth Brooks Dale B. Dalrymple