[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Sheflug] I hate Windoze, Samba and RAID.



>>>>> "Al" == Alex Hudson <eah106 [at] york.ac.uk> writes:

    Al> Of course [using file extension] is autodetection.

To me, detection implies looking at the file contents.  _Users_ have
access to the directory via mv among other things.

    >> It would be rather easy and very efficient to autodetect the
    >> kind of corruption you're talking about: just XOR the high bit
    >> and puke on TRUE.  I don't know why servers don't do that.

    Al> They don't do it 'cos it's time-consuming, I guess.

Less CPU-consuming than packing 8 7-bit bytes into 7 8-bit bytes.  Or
check-summing the TCP packet headers.

Or re-transmitting the broken file in full, with correct settings this
time.  And it would have no impact at all on transmission time on any
link slower than the PCI bus.

    Al> And they don't want the possibility of the user spending
    Al> minutes ftping something only to be told "work done thus far
    Al> is useless, change to binary please".

Isn't that what just happened to Barrie?  Not to mention that not only
was the work useless, it is now irreproducible.  The only thing is
that the server lied (by omission) about the work being done and
useful.  :-(  Furthermore, if the client and server both support
restart, all the work done to that point, save the last packet-ful, is
still valid and can be appended to!

    Al> What does the spec. say?

There's no spec I know of for clients.  But "safety first" is a
sufficiently obvious motto.  Maybe I'm just gullible, but I tend to
give the community credit for doing it that way for the right reason.

    >> With the exception of things like LILO, which is NOT part of
    >> Unix anyway (it's a BIOS app),

    Al> Which bit of it? After updating lilo.conf, I run 'lilo'. These
    Al> seems to me to be a Unix app.... ?

It runs under Unix, sure.  That doesn't make it "part of Unix" any
more than isapnp or executable MP3s[1] are.  If it were "part of
Unix," it would run on Sparcs and Alphas and PPCs, but it doesn't.

    Al> Which kind of puts it at the door of the designers - some
    Al> software designers, particularly under Unix than other OSes,
    Al> have no interest in stopping the user making obvious misakes,
    Al> which is quite sad.

You mean like typing a term paper into Emacs's *scratch* buffer and
hitting C-x C-c, thinking they'd get asked to save?

    Al> I can't think of an example off the top of my head, and I
    Al> certainly don't claim they are anything but a small minority,
    Al> but there are people who take the attitude "Well, if they do
    Al> that, it's the user's fault" which is one of the crappest
    Al> excuses I've ever heard.

I can't think of any examples at all.  Except for all software vendors
(read the NO WARRANTY section of the GNU GPL recently?).

I do know people who say "RTFM" in response to a sad tale of data
loss.  I do that myself (eg, in response to the term paper story).

You know how to make a building 100% safe against loss of life in a
major earthquake?  Fill in the hole.  You know, the big one between
the four walls.  There is always a tension between safety and
functionality, if only in terms of designer and coder time.



Footnotes: 
[1]  Exercise for the reader.  Ain't I evil?

-- 
University of Tsukuba                Tennodai 1-1-1 Tsukuba 305-8573 JAPAN
Institute of Policy and Planning Sciences       Tel/fax: +81 (298) 53-5091
_________________  _________________  _________________  _________________
What are those straight lines for?  "XEmacs rules."
---------------------------------------------------------------------
Sheffield Linux User's Group - http://www.sheflug.co.uk
To unsubscribe from this list send mail to
- <sheflug-request [at] vuw.ac.nz> - with the word 
 "unsubscribe" in the body of the message. 

  GNU the choice of a complete generation.