[ Fabrizio Oddone | Italian Recipes | Ricette Italiane | Register My Shareware | Y2K | Site Map | What's New ]

Mac OS X Info

What's this?

A new idea for the Internet. Technically speaking, a new Uniform Resource Locator (URL) scheme. Besides the specification for the fab URL scheme I am going to supply the first implementation, a "helper application" (tentatively dubbed QuickestMirror) for the new scheme.

Purpose

How does it work?

fab://ftp.stairways.com/stairways?

QuickestMirror will show an ordinary list like the one shown below. The list is dynamically sorted so that the faster sites drift toward the top.

On Sunday, Nov 2, 1997 I read this on MacInTouch:

The freeware Eudora Light 3.1.3, a new version, is posted on a download page, but we had problems with the FTP server this morning.

Here is where QuickestMirror comes useful. Just try fab://ftp.eudora.com/eudora?eudoralight/mac/english/3.1.3/ to witness.

How does QuickestMirror know the host list?

It is certainly possible to devise a protocol allowing to retrieve such lists dynamically. Interested parties may want to know about spoofing attacks first.

Will my favorite Internet app take advantage of the new URL scheme?

There are notable exceptions, of course, such as Netscape: the popular browser requires an AppleScript kludge like this:

tell application "Netscape Navigator"
register protocol "QMir" for protocol "fab"
end tell

This instructs Netscape to let QuickestMirror handle fab URLs. In theory you should run this script only once, but...

I've just learned that Netscape Navigator, versions 3.0 through 3.0.2, sometimes cannot retain what protocols have been registered. More information about this can be found at the Netscape Preferences Fix home page. You can download the fix itself from the Info-Mac archives, just 9K.

How do you pick a server?

This and other important results are found in this most interesting paper by Mark E. Crovella and Robert L. Carter, which I found almost by chance with Alta Vista.

Since a complete understanding of this paper requires a background in maths and statistics, I summarize here their findings (these sentences have more or less been excerpted from the paper):

• The number of hops between two Internet hosts is not strongly correlated to round-trip latency. Thus, the distance in hops between two hosts is not necessarily a good predictor of the expected latency of a document transfer.

• The extra cost at runtime incurred by dynamic latency measurement, as compared to prior static knowledge of hop distances, is well justified based on the resulting improved performance.

• Selection based on dynamic latency measurement is preferable to random selection.

• The effect of randomly distributing replicas throughout the Internet is to sharply decrease the latency between a client and server; that is, you do not need to devise a host replication strategy. As an added bonus, you gain fault tolerance.

• Network congestion plays a significant role in the overall distribution of round trip times.

• Dynamic policies consistently outperform static policies; furthermore, the difference between static policies and dynamic policies increases with increasing document size. In fact, even random server selection is preferable to choosing any static server for documents larger than 5K bytes.

Requirements

QuickestMirror ships as a "fat binary" application and requires a 68020 processor. If there is demand, I will build a 68000 version for separate distribution.

Things that you won't find in the readme docs

• QuickestMirror measures Accurate Round Trip Time as described by Sustainable Softworks.

• Marco Piovanelli reports that QuickestMirror fails behind a proxy server, and Karl-Johan Norén explains why and has good suggestions for the next release:

  I just realised a problem with this scheme, though (which is a real problem inherent in it): proxy servers. A proxy server is likely to prohibit using more than four simultaneous connections from the same host, maybe even as few as two (you'll have to check the HTTP 1.1 draft for details). It might thus be a good idea to be able to set preferences for the order in which the alternative hosts are to be contacted, otherwise the first four listed alternatives might be over the Atlantic, but there were European alternates that were stopped by the proxy.
  
  Also, in cases where you have a very good connection with an alternate (like in my case -- the distance from this computer to ftp.sunet.se is less than 10 km, and the slowest part is my department's 10 MB/s Ethernet, the rest is all 100 MB/s), you should be able to configure fab to always use this alternate and only fall back to the list if this one refuses the connection.

• Monish of the Macintosh News Network had a few interesting comments:

  Is there a way (perhaps an option) that QM could continue in the background without becoming the frontmost app; i.e., after clicking on a fab URL, it would launch in the background rather than bringing the app in foreground as it did in my preliminary testing?

  As a matter of fact, IE kicks QM to the foreground, whereas Netscape leaves QM in the background. (At least this is what happens here, assuming that QM is already running!) Indeed, this could be added as an option, but if IE forces QM to the foreground, an ugly double-switch would occur.

  For reasons I don't quite understand, Microsoft Internet Explorer 3.0 will download ftp URLs using its Download Manager (even though the ftp helper is Fetch), but using QuickestMirror you are correctly pushed over to Fetch given the helper configuration in InternetConfig. Is there (or should there be) an option to return to the browser for ftp, if necessary? I realize that this is contradictory, but I never thought about the fact the IE continues to do ftp despite the IC configuration for Fetch.

  Ed "What the" Heckman found the answer to the riddle!

  IMHO, Internet Explorer has an advantage over Netscape in that it actually supports Internet Config. Its support is overridden by a checkbox in IE's preference settings for the helper applications labelled "Use current application if possible". In essence, this means "Ignore Internet Config". And interestingly enough, it is checked by default. So it's yet another case of Microsoft's "Embrace and smother... uhhh... extend" techniques. But it still beats Netscape ignoring IC completely.
  
  IMHO, I would have to say to stick to how IC is configured. If they really want IE to handle the FTP downloads, they can update IC to say so. That would be much simpler (and smoother) than having to try and figure out what that checkbox is set to.
  
  BTW... I hope I haven't insulted you by stating something you didn't already know. :-)

  I'm always ready to admit my ignorance. :-)
  To access the setting, go to Preferences:Protocol Helpers , then double-click an entry in the list.