A new definition of libraries (was: Linux Installation)

[Bob Warren]

Some people might think I am flogging a dead horse. In my opinion, the 
horse is still very much alive, and this fact has enormous consequences 
for Rev Linux users, not only in theoretical terms but also in terms of 
practices.

According to what has been discussed under the thread "Linux 
Installation", the answer to the question "When is a standalone not a 
standalone?" is "When it is a Rev application".

I have had all kinds of explanations about the history of the term, 
reminders of my lack of knowledge about what really happens internally 
in an OS, etc. However, I find myself in a position where the whole 
business is still as clear as mud, and I don't know what to do in terms 
of the distribution of my Rev applications for Linux.

Let me try introducing a new concept. The terms "constant" and 
"variable" are familiar to us all, but have you ever thought of applying 
  such terms to libraries? According to this idea:

a) "Constant" libraries are contained within an OS (in the case of 
Linux, probably within the kernel). They are responsible, for example, 
for the everyday I-O of reading/writing to the file system, and so on.
In the case of Linux, such constant libraries are common to all distros.

b) "Variable" libraries within an OS are the ones that can change from 
time to time, or can be different between one distribution and another.

When I tried to define "standalones" previously, I said more or less 
what I am going to say now, but without the new terminology. Modified 
with the terminology, we now have the following hypothetical definitions:

"Standalone", used to mean what it says rather than in relation to its 
historical X-Talk significance, means:

EITHER:
1) A program which in no way refers to any type of library within the 
runtime OS

OR:
2) A program which depends on "constant" libraries in the runtime OS only.

To these 2 types of program we need to add non-standalones:
3) A program which makes use of "variable" libraries in the runtime OS.

I had hoped that Rev "standalones" belonged to category 2, not therefore 
requiring any kind of "installation" in the normal (e.g. Windows) sense 
of the term. People tell me theoretically that they belong to category 3 
and that they therefore do require a full-blown installation.

Since I am not essentially a religious kind of person, I try very hard 
to believe what people tell me, but unfortunately I need to see things 
with my own eyes in practice in order to be 100% convinced. So far, 
nobody has told me of a Rev standalone that cannot be transferred 
successfully to other distros of Linux (of course, leaving aside the 
question of different references it might make to the different 
characteristics of the particular file system). Only time will tell 
whether Rev standalones really belong in one category or another by this 
empirical method.

Of course, one person who should know about this better than any other 
is Mark Waddingham, Rev's chief technical officer. Now, an amusing 
picture conjures itself up in my imagination. There's Kevin with a whip 
in his hand and there's Mark with beads of sweat on his brow doing his 
daily debugging. And Mark gingerly raises his right hand and mutters, 
"Can I go for a wee wee-wee?". And the stern reply is "Later!". So I 
won't send any e-mails to Mark, but it would be nice (if ever he gets 
time to read the UR-List) if he could give us a position as to what the 
status of the Rev "standalones" (especially in Linux) is supposed to be.
Or perhaps somebody could jump on him at RevCon? How about you, Richard? 
  It's your thread (I didn't steal it - just borrowed it for a while)!!

Bob