URIs and URLs are not quite the same thing. URL is a special kind of URI that specifies a physical server that a document exists on. There are some protocols (such as Freenet (external link)) where a peice of data exists "out there", but you can't pinpoint an exact location. In such a case, you can't use URLs, but URIs work just fine.

Anyways, URI syntax for files is just 'file://' followed by the absolute path to the file. So an example of a Win32 file URI would be 'file://C:\windows'. On *nix, you get a horrid series of toothpicks: 'file:///usr/local' (though I've noticed that KDE assumes the begining slash on the filename, thus allowing 'file://usr/local').

Update: I forgot that ':' has special meaning in a URI. As another poster pointed out, a Win32 file URI should be 'file://C|\windows'.

----
Reinvent a rounder wheel.

Note: All code is untested, unless otherwise stated

Last I read, URL is being dropped from technical documents. At http://www.w3.org/Addressing/ "URL Uniform Resource Locator. An informal term (no longer used in technical specifications) associated with popular URI schemes: http, ftp, mailto, etc." Also http://www.w3.org/Addressing/9710-uri-vs-url.html.

I find what you mentioned in the 1998 spec, and I guess the "cool URLs don't change!" backlash drove the quest that things on the web be URI's, not something that happens to be the current location but might change.

They generally contain three slashes, because you take a file path that starts with a slash (such as /etc/passwd) and then add the file:// to the beginning, so you get file:///etc/passwd. The vertical bar is what windows uses in place of :, since the : has a different meaning in a URI, so C:\foo.txt becomes file://C|/foo.txt.

Mozilla is showing three slashes before the drive letter, meaning a leading slash before the drive letter. Hmm, and it's showing the colon as-is! IOW, file:///C:/foo.txt.

So, if | is used for :, what's the meaning of : within a URI? I know it's used to separate the original service name from the rest, but don't recall any special meaning on the right side. Is that true for all colons in a file name (on systems where that is a legal part of a name, or has a special meaning to the file system)?

And what do you do about |'s in the name that were there to begin with?

—John

So, if | is used for :, what's the meaning of : within a URI? I know it's used to separate the original service name from the rest, but don't recall any special meaning on the right side.

I don't think it has a special meaning, but it's a reserved character for the entire URI. And yes, many browsers will let you get away with using ':' instead of '|'.

And what do you do about |'s in the name that were there to begin with?

You beat up the person who named a file with a '|' (:

----
Reinvent a rounder wheel.

Note: All code is untested, unless otherwise stated

The reason there are three slashes is that there is an optional host name that comes between the second and third slash: e.g. file://novell01/departmentshare/me/myfile.txt

I read that Perl 6 is going to avoid the filename syntax issue by using URI's (also known as URL's) to provide a platform-independant file naming convension.

Yipes!

I can't find it in the Apocalypse now. But I know I read something quite specific on that idea, that unifying file names across OS's by using URL syntax will eliminate some of the crud in writing portable scripts. Perhaps it didn't make it into the Apoclypse, or at least not covered yet.

—John

I think it would've made it into the apocalypses already. You may have remembered something from perl6-language or one of the RFC proposals, since I recall it being proposed, but it's definitely a non-starter. Optional, perhaps, but definitely not mandatory.

I read that Perl 6 is going to avoid the filename syntax issue by using URI's (also known as URL's) to provide a platform-independant file naming convension.

Really!? What a wonderful way to alienate the entire Windows world. The decision to reserve colons in URI was a bad idea. Its the reason that Win32 folks dont like URI's. To embrace it in perl6 is an even worse idea.

I know that this is a pretty negative POV, but I'm willing to bet that whomever thought up borrowing URI's for perl doesnt use a Win32 box much. file:///D|WTF!

Oh well, I suppose ill get used to it at some point. *sigh*

About the use of file://c:/foo instead of file://c|/foo, weel I think that the both will work, since we have many programms that use them today.

About the use of reserved characters in file names, well, we can't tell to someone that use a OS to not use some symbol because in other doesn't work! In other words, in theory we need to can put "any" symbol in file names. To do that, how about use the %HH, that every body know in the URI? We already use that for spaces, %20.

This is a inportant point, since when we talk about Perl6 we say Parrot, and Parrot is not only for Perl, is a VM for many languages. And I think that some upgrades to the rolls of URI can be made, since URI was defined a "long time ago" and was tested very well, and now we can know the best options to choice.

Graciliano M. P.
"The creativity is the expression of the liberty".

In the URI::File module, it shows

  Unix                  URI
  ----------            ------------------
  foo/bar      <==>     foo/bar
  /foo/bar     <==>     file:/foo/bar
  /foo/bar     <==      file://localhost/foo/bar
  file:         ==>     ./file:
  <undef>      <==      file:/fo%00/bar
  /            <==>     file:/
[download]

that is, one slash only if no host, but doesn't show three slashes for that same case. But, I thought the leading // after the file: meant that the rest of the name uses the slash-separated hiarachy, otherwise is opaque, so one slash isn't right? But that wouldn't explain relative URL syntax.

If we have %HH syntax, why map : to | at all? It's not like anyone types filename urls in that syntax.