Posted by Jerry Stuckle on 03/09/07 05:57

Malcolm Dew-Jones wrote:
> Jerry Stuckle (jstucklex@attglobal.net) wrote:
> : Malcolm Dew-Jones wrote:
> : > andy.z@blueyonder.com wrote:
> : >
> : > : If 2 people try to access the same text file at the same time to write to
> : > : it - what happens in PHP ?
> : >
> : > : What I mean is - presumably the first will be ok -
> : > : But what will the second person actually see in his browser?
> : > : And what will php do about it if not specifically programmed to deal with
> : > : it
> : >
> : > : ie whats the default behaviour?
> : >
> : > At this point it may be useful to know that if you open a file in _append_
> : > mode on unix/linux then each write statement writes correctly as a single
> : > "unit" at the _current_ end of the file, even if multiple programs are
> : > writing to the file and continually changing the size of the file.
> : >
> : > That means that if multiple processes all write one line at a time to
> : > something like a log file then the log file grows one line at a time.
> : > The lines will be in the order they were written, but will not otherwise
> : > interfere with each other. Indeed this is how a basic unix log file
> : > works.
> : >
> : > HOWEVER that only helps with simple text files where you can append things
> : > at the end. If you have something like an HTML file then the updates have
> : > to happen in the middle of the file (before the </body></html> for
> : > example.
> : >
> : > In that case you need to lock the file before you modify it. You have to
> : > do that yourself. There are lots of examples of what you have to do so I
> : > won't bother trying to remember the exact details here and now.
>
> : This is only true if there is no buffering involved.
>
>
> : If there is buffering (as there almost always is), anything can happen.
> : If for instance, one script tries to write 800 bytes with a 512 byte
> : buffer, the first 512 bytes (only) may be written. And the second
> : script may get in between the first 512 bytes and the remaining 288
> : bytes to write its data.
>
>
> : Even worse - to append to a file, the system must read the file into the
> : buffer. If this takes 500 bytes, the next 12 bytes can be written - and
> : the rest buffered. Then the second script might put data after the
> : first 12 bytes but before the remaining data.
>
> man open(2)
> (snip)
> O_APPEND
> The file is opened in append mode. Initially, and
> before each write, the file pointer is positioned
> at the end of the file, as if with lseek.
>
> O_APPEND may lead to corrupted files on NFS file
> systems if more than one process appends data to a
> file at once. This is because NFS does not sup-
> port appending to a file, so the client kernel has
> to simulate it, which can't be done without a race
> condition.
>
> Corruption on * NFS * is explicitly mentioned because that is * not * the
> normal result of using append mode on unix/linux (until a system wide
> limit is reached which is on the order of 30,000 bytes).
>
> The file routines of other languages seem to have no problem with multiple
> processes using append mode against the same file, though I suppose that
> the php file routines could do something that inteferes.
>

Keep believing that. But don't come running here when you get a
corrupted file. It happens. The time gap is very small (unless the
server is very heavily loaded). But it does exist. And it exists on Linux.

--
==================
Remove the "x" from my email address
Jerry Stuckle
JDS Computer Training Corp.
jstucklex@attglobal.net
==================

[Back to original message]