10 ms·

My direct personal experience - as attested by the guys I work with - is that upgrading something like Python, or a browser, breaks my machine for up to a week

My direct personal experience - as attested by the guys I work with - is that upgrading something like Python, or a browser, breaks my machine for up to a week as I fight through dependency hell. The machine is running SuSE 8.1 with a broken YaST. Upgrading Python - I suspect - would require upgrading all the C libraries, etc. Easier just to wipe that machine and start from scratch, but why bother? This was just a trivial play on the side.

jacquesm 16y ago
chroot is your friend in such cases.
- RiderOfGiraffes 16y ago
  But then I would never have had the fun of exploring these different versions of a classic algorithm. There's a bigger point about your comment though. I've learned through painful, painful experience to be wary of people who say "Why not just ..." It usually hides a weath of knowledge and experience that they take for granted, but which takes time to accrue. Upgrading Python might just take half a day, or it might take a week. How do I know in advance what that would be? How do I know that in the middle of it I'll still have a usable machine at all for stuff I need to be doing? There's the old saw about stopping to sharpen the axe because, in the long run, it'll be quicker. I'm always very aware of that. I've found that trying to "sharpen the axe" by way of "upgrading" my machine always ends up costing far more time than it saved. Your experience will no doubt be different, but your knowledge base is different from mine. There are swathes of things I don't know, don't have time to explore, and don't really care about. In truth, I wrote the versio that blew the stack, played around for 30 minute because it was interesting, took a break, then wrote the obvious final version. Time well spent, and certainly more fun that upgrading Python. (On that machine).
  - jacquesm 16y ago
    Oh, I agree about the fun part :) But when you're optimizing something that usually indicates you would like to have your results faster! ;) Dependency hell is one of the bigger shortcomings of the way software is distributed, my way of dealing with that is to link against static libraries or use that chroot. For most packages static linking is a compile time flag. Typically though, compiling from source should work, it's when you start upgrading from binaries in older repositories that the trouble starts.

jacquesm 16y ago

chroot is your friend in such cases.

RiderOfGiraffes 16y ago
But then I would never have had the fun of exploring these different versions of a classic algorithm. There's a bigger point about your comment though. I've learned through painful, painful experience to be wary of people who say "Why not just ..." It usually hides a weath of knowledge and experience that they take for granted, but which takes time to accrue. Upgrading Python might just take half a day, or it might take a week. How do I know in advance what that would be? How do I know that in the middle of it I'll still have a usable machine at all for stuff I need to be doing? There's the old saw about stopping to sharpen the axe because, in the long run, it'll be quicker. I'm always very aware of that. I've found that trying to "sharpen the axe" by way of "upgrading" my machine always ends up costing far more time than it saved. Your experience will no doubt be different, but your knowledge base is different from mine. There are swathes of things I don't know, don't have time to explore, and don't really care about. In truth, I wrote the versio that blew the stack, played around for 30 minute because it was interesting, took a break, then wrote the obvious final version. Time well spent, and certainly more fun that upgrading Python. (On that machine).
- jacquesm 16y ago
  Oh, I agree about the fun part :) But when you're optimizing something that usually indicates you would like to have your results faster! ;) Dependency hell is one of the bigger shortcomings of the way software is distributed, my way of dealing with that is to link against static libraries or use that chroot. For most packages static linking is a compile time flag. Typically though, compiling from source should work, it's when you start upgrading from binaries in older repositories that the trouble starts.