Re:Tmpfs considered harmful
by rwmj on Wednesday September 12, 2012 @09:53AM
(#4510639)
This is exactly the point of the posting. Every program has to be examined for this problem, and it forces a complex decision about the storage hierarchy onto every programmer.
[ Parent
| on HN ]
Every program has had to be designed with this in mind for decades or be broken on any number of systems. This may be new behaviour for Fedora, but it's not a new thing on Linux, or for Unix-like systems in general.
It's not just systems that delete on reboot, but also systems set up with cron-jobs to regularly wipe /tmp, or sysadmins that will (rightfully) see /tmp as something that can be wiped at pretty much time.
That's like saying "RAM considered harmful", because every time a developer has to think "do I just store this in memory, or do I sync it to disk" "it forces a complex decision about the storage hierarchy": /tmp is simply RAM that has the API of a file; it used to be implemented using a horrible kludge with files on disk that were periodically deleted, but with tmpfs the dream became a reality.
/tmp is simply RAM that has the API of a file
...which used to behave, physically, as disk, and now behaves as RAM. If I'm doing large file manipulations that require temporary files larger than RAM where should I put them?According to the comments on the original article, that location is /var/tmp. Apparently it's some sort of convention/standard.
First time I've heard about /var/tmp.
Re:Tmpfs considered harmful
by regularfry on Wednesday September 12, 2012 @04:07PM
(#4512853)
/tmp. If it's tmpfs, it can expand into swap space.
Swap is usually no bigger than two or three times your RAM. Disk in general is 10 to 100 times the size of your RAM.
You should take all of the space you previously felt acceptable to provide /tmp and allocate it to swap space. The history and semantics of that folder and how it relates to RAM, especially on systems that could not implement swap space, actually make this make sense: if you upgrade to tmpfs from "periodic deletion of /tmp" and are not increasing the size of your swap partition accordingly, you are doing something wrong.
Re:Tmpfs considered harmful
by regularfry on Thursday September 13, 2012 @03:37AM
(#4515156)
If you know you're going to need more space, you can just create a transient swapfile for the duration. Or use /var/cache, if you suspect you know better than tmpfs and the swap allocator.
Re:Tmpfs considered harmful
by prodigal_erik on Wednesday September 12, 2012 @12:22PM
(#4511407)
RAM really is a harmful kludge. Neither Turing machines nor lambda calculus call for some of an algorithm's state to suddenly be obliterated. The only reason we put up with it is that disk and SSD are so much slower.
Speed is not the only benefit of RAM. The internal data structures of most programs are gnarly messes of intertwined pointers, fragile, easily corrupted, hard to upgrade in place, almost impossible to fix when they break. The division between RAM and disk forces everything to be serialized at intervals to clean storage formats that can be reread for a fresh start. As a practical matter, that's a big enough benefit to justify enforcing the division at an API level even if disk were as fast as RAM.
Re:Tmpfs considered harmful
by prodigal_erik on Thursday September 13, 2012 @12:43PM
(#4517153)
The industry might produce less worthless crap without the crutch of "it's not quite unusable if you start over often enough".
I've never understood the obsession with Turing machines. It's like basing one's core belief of what a car is off one of Da Vinci's crazy wind-up things.
Ram is just state. Turing machines require a non-tape state machine mechanism. You could consider ram a part of that, 4 gigs corresponding to many many possible states.
Some people (I do not agree with this view) believe that applications should be universally persistent: so that they never "stop running" and never "lose data" even when the machine restarts or even if the programs are transferred between systems.
Do you consider distros that rm -rf /tmp on startup to be broken? If not, why not?
Firstly Fedora (<= 17) doesn't do this. It has a tmpwatch cronjob which deletes files after 10 days (/tmp) and 30 days (/var/tmp).
Secondly, no I don't think it's broken. But it doesn't in any way affect the fact that as a last resort I can reboot in single user mode and grab the precious mutt file before it gets deleted.