[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

RE: st: re:r(601) "server says file temporarily redirected to [URL]"

From   "Nick Cox" <>
To   <>
Subject   RE: st: re:r(601) "server says file temporarily redirected to [URL]"
Date   Fri, 24 Oct 2008 17:57:50 +0100

I do agree with your main point, and especially your last sentence, but
in no sense is it controversial. And I want to re-emphasise mine. I
think there is a big difference between 

1. Why did I lose my statistical results from -foo-? Oh, I see: because
I ran another statistical command of similar kind afterwards that
overwrote the results. 


2. Why did I lose my statistical results from -foo-? Because I ran a
command that accessed the internet afterwards! 

1. should make sense to users, but 2. I think would, and should, seem
more unreasonable. 
I believe that StataCorp would not implement 2.; it is important that
users don't either (in so far as they can). 


Sergiy Radyakin

I leave the implementation details up to the higher authority
realising that there is always a possibility to declare it n-class
(for network commands) or return results in a global (similar to
S_MACH or S_OSDTL) so that it does not bite anyone.

As for changing the class of commands, I think it is up to the
programmers to write robust code and not rely on the current
declaration of commands. E.g. if -foo- is rclass and -bar- is not, it
is not robust to write:

foo arguments1
bar arguments2
// here analyse the r()-results of foo

because in the future if bar becomes r-class (or utilizes r-class
commands within itself) it will destroy foo's results. A good practice
is to retireve saved results immediately after the command that
produced them.

On Fri, Oct 24, 2008 at 9:57 AM, Nick Cox <> wrote:

> Making -net- r-class would necessarily mean that net access would zap
> the results left behind by any other r-class command, quite apart from
> any criteria for what an r-class command should be. I think that would
> bite far more people than it helped.
> I can see a case for leaving something in memory, but it would be
> (say) c-class.

Sergiy Radyakin replied to Kit Baum

> thank you, -wget- worked fine. I tried -telnet- before, but it was not
> redirecting either.
> I agree to the security provisions, but if the URL2 were accessible
> for programmers, then I could decide within my program whether to
> follow it or not. Since Stata currently extracts that new URL from the
> server's response, I believe there is not too much cost of saving it
> to an r() value. Relying on an external utility for this purpose
> limits portability of the resulting .do file, and adds to testing,
> maintanence and deployment headaches.

*   For searches and help try:

© Copyright 1996–2024 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index