Slash/Backslash\etc was: Voice-over Flubs

[Ric Werme]

I haven't followed all the Email on this thread, but what the heck.

Dan Strassberg wrote:
> Well, I can't say that I carefully examined the contents of IE's
> destination field after I entered the URL with the backslashes instead
> of slashes. I think, though, that I did notice that the backslashes
> had been transformed to slashes.

This would be a sensible thing for a Windows system to do since it
doesn't use backslash for much of anything except the path name
separator.

> ... But since rerouting by the server
> also--apparently--modifies the contents of the destination field, I
> don't see how a user with no special insight into what happens in the
> domain-name server could guess that the browser had performed the
> trick.

Ah, but geeks mit tools haff our vays....

All the domain name server does is take a query for a host name (e.g.
www.google.com) and returns the numeric IP address.  For Google, there
may be many as that's one way to spread the load:

  $ host www.google.com
  www.google.com is an alias for www.l.google.com.
  www.l.google.com has address 64.233.169.147
  www.l.google.com has address 64.233.169.104
  www.l.google.com has address 64.233.169.99
  www.l.google.com has address 64.233.169.103
  www.google.com is an alias for www.l.google.com.
  www.google.com is an alias for www.l.google.com.

For me, my ISP doesn't have to bother:

  $ host WermeNH.com
  WermeNH.com has address 64.136.25.175
  WermeNH.com mail is handled by 10 mail.bluetie.com.

  $ host 64.136.25.175
  175.25.136.64.in-addr.arpa domain name pointer public-25-175.lax.ws.untd.com.

For a Web browser, no slashes go to the DNS server.  Backslashes can, but
it looks as though Firefox on Linux took the URL
http://werme.nh.com\sdd\index.html, picked out "werme.nh.com\sdd\index.html"
as the server name, noticed the illegal characters in it and replied
"Firefox can't find the server at werme.nh.con\sdd\index.html" without
bothering the DNS server.

Browsers have to pick out protocol name (e.g. from "http:"), server name
(e.g. from //werme.nh.com) and the rest (/sdd/index.html).  For that URL,
Firefox opened a connection to host 64.136.25.175, port 80 and sent:

  GET /sdd/index.html HTTP/1.1
  Host: wermenh.com

and a bunch of other hints and offerings, then gets the index page back.

> I assume that IE does this magic only to backslashes entered into the
> destination field. Oh, and IIRC, the transformation doesn't take place
> until you click on "go," which enhances the impression that something
> outside the local PC is making it happen.

IE would have to accept backslashes to pick out the hostname, so it would
be completely sensible to change the path (e.g. "sdd\index.html" to
"sdd/index.html").

Given the number of people who tell me to type backslash when they mean '/'
I've all but given up correcting them.

-----

Garrett Wollman wrote:

> It's your browser that does that, not the server -- the server never
> sees the backslash.  This behavior originated in IE but has
> apparently
> spread to other browsers recently (Opera does it; I did not check
> Firefox).

I did not check Firefox on Windows, it may.

> I blame Microsoft for using the wrong path separator character.

That goes back to the early 1980s.  MS-DOS was largely a port of
CP/M, which was influenced by various systems from DEC in Maynard
dating back at least to the PDP-6 in the 1960s.

Those systems used forward slash to note options, e.g. to copy a file in
binary mode, one might do:

  .R PIP
  * NEW.FIL=OLD.FIL/B
  * ^C
  .

After a while, the PDP-10 (successor to the PDP-6) did support one
directory per user, but the syntax was something like DSKB:[10,417]NEW.FIL
wherel [10,417] was my project-programmer number.  Meanwhile, AT&T went
off and created Unix, and things began to look like
/dskb/users/werme/new.file.  Unix and Linux still use that, though on
my system, it would likely be in /home/werme/txt/new.file.

Microsoft used to have their own PDP-10s and that helped influence a
lot of their work.  They eventually figured out that the Unix pathname
idea made a lot more sense and kind of adopted it, but kept the
device name part.  Worse, since they were using slash to introduce options,
they had to do something different and either some committee or some
programmer working at 0200 picked the obvious quick and dirty hack of
using backslash.  Apparently MS-DOS and maybe windows can be configured
to use a different separator, but I don't know if the system would remain
usable if someone were to use slash.  Behind those mouse clicks are
little files specifying programs, filenames, and options that use slashes.

At any rate, now my file on Windows looks something like
"D:\Users\Werme\Documents\New file.doc", gets examined by virus checkers
a dozen times a day, and annoy geeks who know where it came from.
It would have been so much better if they had just done the hard work and
adopt the Unix format.

Exacty how someone can think a line tilted forward should be called
backslash is beyond me.

      -Ric Werme