@acer 

Is your understanding that similar considerations hold for externally called routines as for kernel builtins? I would guess yes, but ?timelimit is even more obscure on this case.

@tomleslie 

You may like to read this thread on performance issues in Mathematica and Maple. In particular, on seq and `$` acer stated there:

Jacques writes, "..why do the designers make it so darned easy to write bad code?" One partial answer is that it is misplaced subservience to users' "wishes for convenience". People sometimes want something easier than `seq`, and are given `$` whose difference they often don't understand and get into difficulty.

I make the observation that people not understanding the difference may include support. And a question: where is this important difference clearly documented?

@tomleslie 

Do you mean this one? I do not see any answer in this thread. In any case, after ?seq and ?$, both forms are described as thread-safe.

@Carl Love 

But then the complementary set of the set {x>4,y>4} in the plane (x,y) is not {x<=4,y<=4} (only).

How is "supplementary set" defined?

@acer 

Thank you, that saves time. By the way, about the cited syntax difference between 1D and 2D views for that help example, even with the former (my default) the sentence before still typesets in the the 2D form like "Adding filters out one of these solutions". I find this syntax discrepancy confusing for the ordinary user.

@tomleslie 

Yes, frequently browser issues have been described. I have many examples of them in my archive. For instance this thread, actually incomplete, missing  several messages describing additional browser problems (I keep the complete version).

@Markiyan Hirnyk 

These error messages state what the system was doing at the expiration time. As there may be variations in the sequence of steps and/or the time when they occur, the system may be catched doing different things in successive runs, as you have got. In the second one, it is caught while reading the binary member "file" solve.m of the repository maple.mla, see e.g. ?repository and ?Formats,m . The offset number is the position within the binary file where the reading stopped, presumably.    

@tomleslie 

I doubt that you could say much of my programming practices, hence the difference with yours. Note, in particular, that I have said nothing about them in my comment above... Nevertheless, FYI, my programming preferences are much more close to yours than you may believe :)

The interesting thing is that Maple's library shows many examples of procedures violating the guidelines that you have stated. (and actually those stated in Maple's documentation as advice to the users). E.g. procedures of over 300 lines of code, packed with conditional statements, quite hard to debug indeed.

@tomleslie 

This Matlab's document Scripts and Functions, describes briefly Matlab's programming language, and it states that "internal" variables are recognized local to the function (automatically, without any warning). And probably, these two different languages allow/promote different programming practices. So, it does not seem to me fair to blame this way someone used to program in such a different environment.  

@Carl Love 

In a short search I have came across with the project PURRS: The Parma University's Recurrence Relation Solver. It mentions "some multivariate recurrences". Anyway, for linear multivariate recurrences as in this example, it seems like the ansatz M(k,n)=lambda^k*mu^n does work.

@Carl Love 

In my experience, the same code may run O(10) times slower in Standard wrt the CLI. Not to speak of Standard's tendency to crashing (despite lots of RAM), worksheet corruption, etc. So, in my opinion, and for my usage, Code Edit Regions or other features of Standard are, typically, not worthwhile tradeoffs.

@ 

You may be interested in reading e.g this thread on Classic, if you did not already. In short, Classic was at a technological dead end street already a decade ago. So, the true dichotomy is not Classic vs Standard, but a single GUI (Standard, say) for everybody vs specialized GUIs for users with different needs. Laurent's choice was for a single GUI, as stated in the linked thread. I find pretty clear that his choice was wrong (independently of the serious design problems of Standard itself).

@tomleslie 

I would add to your comparative review that these are two different languages, with different syntax, parsers and error messages. And these differences are not fully explained in the documentation. And the aesthic prettiness of the 2D language comes at additional costs that the user should take into account: e.g. running code in 2D language is more inefficient, and the 2D parser is more buggy, compounding with the aforementioned difficulties in debuging.  

@Markiyan Hirnyk 

Yes, it may be a difficult matter, but what matters, in my opinion, is the priority given to this issue. E.g. it looks like it is much higher in the Sage project:

Sage, LaTeX and Friends

LaTeX in the Cloud