Boris Zbarsky wrote:
Sorry about the delay, the real world intruded.
> Where do names appear? It's designed by someone who assumes that given
> an object you can get meta-information about its type from the OS
> (which you can on sane OSes; see BeOS).
BeOS is just one OS. Mozilla is supposed to be multi-platform and names
are relevant on some platforms.
> In 99.99% of cases, the file type will in fact not be changing any time
> soon. Let's not call things "useless" when they are at worst slightly
> unreliable.
Oh, let's. If treating the common case were the only requirement then
mozilla would have a lot fewer tests for null this and negative that. I
would think everyone would want mozilla to be a robust product. Given the
possibility of virii and worms (off topic, I think it's possible to infect
mozilla), I would be very paranoid, particularly when uploading.
Also, "useless" was referring to opening a bunch of files just to see
what's in them. Also (again), "accept" support seems to be optional so I'm
not sure this is something worth pursuing at this time.
> Realistically, in all cases I deal with, the chance of a power outage
> during the file upload is higher than the chance of a file being
> overwritten in the time while I am selecting files to upload. Perhaps
> the files on your system all get replaced a lot?
I have lots of files that are changing. They're mostly log files but I can
open any file for reading while it's also open for writing. It's usually
hard to detect this situation. I wouldn't want to presume what users'
might want to upload.
Then there are the changes I've made by mistake, or the cron job that
purges temp directories, or other possibilities.
As I already said, the system is mutable, mozilla should make no (or very
few) assumptions about what's happening.
> There would be no "removing special cases" involved. File upload has to
> do special things with the file no matter what (prepend various headers
> to it), so here would be no improvement in perf or footprint of any kind
> involved. In fact, there would likely be a perf loss due to the extra
> copying and function alls.
> Boris
Presumably headers are in a memory buffer so you just stuff that down the
network socket first, then you read from the file stream and just start
stuffing that into the socket. I don't see a problem.
Mozilla isn't a high performance server anyway. If it were it would most
likely have to do special tricks for each platform, e.g. use sendfile on
Linux. Since it's not I don't see network throughput as a big issue.
Besides, in many (most?) cases the network is slower than mozilla anyway
so it doesn't matter.
--
Saturn
2003-08-19 14:34:21.427 UTC (JD 2452871.107192)
X = -0.736742935, Y = 8.304339156, Z = 3.461730550
X' = -0.005857612, Y' = -0.000525271, Z' = 0.000035214
