| 1 |
On Sunday 14 February 2010 20:44:32 Enrico Weigelt wrote: |
| 2 |
> Volker Armin Hemmann wrote: |
| 3 |
> > no, but with static exes you have to recompile everything |
| 4 |
> > everytime a security bug is found. |
| 5 |
> |
| 6 |
> That's the job of the distro buildsystem. Ah, and that dramatically |
| 7 |
> minimizes the chance that things break apart (i still remember |
| 8 |
> the old times when libc updates tended to be dangerous). |
| 9 |
> |
| 10 |
> > Oh - and didn't you just complain about bloat? Nothing means |
| 11 |
> > more bloat than static binaries. |
| 12 |
> |
| 13 |
> As already said, all this under the axiom that libs are *small* |
| 14 |
> and complex/redundant things are done by separate services. |
| 15 |
> Perhaps you might have a look at Plan9 and how its done there. |
| 16 |
|
| 17 |
|
| 18 |
To be fair, Plan9 is Unix done right. |
| 19 |
|
| 20 |
For all it's power, Unix (the system, not just the kernel) has some very |
| 21 |
severe flaws. Why can't I prepend data to a file using any of the common |
| 22 |
shells? Why are pipes 1 input 1 output, instead of the more useful 1 input |
| 23 |
same data to 2 or more outputs? Why is the permission model so simplistic? Why |
| 24 |
is ELF so prone to bloat (or more accurately why do so many compilers generate |
| 25 |
such large libs?) |
| 26 |
|
| 27 |
The answer is because of the available constraints at the time these things |
| 28 |
were introduced. Partly the amount of grunt available from systems of the |
| 29 |
time, partly the speed of disks, partly to keep things simple and to an |
| 30 |
irreducible minimum, with a huge helping of how easy a platform it is to |
| 31 |
develop on. |
| 32 |
|
| 33 |
For better or worse, what we have is what we have and it's the sum total of |
| 34 |
the past. |
| 35 |
|
| 36 |
|
| 37 |
-- |
| 38 |
alan dot mckinnon at gmail dot com |
Archives