News and commentary from the cross-platform scripting community.
Mail Starting 7/19/98
That security risk that was pointed to (collecting personal info) is nothing compared to the idea that some places include the username and password for certain sites in their URLs. If these URLs are then bookmarked, then uploaded to this bookmark databank, there could be some real problems there.
From: email@example.com (Ross Olson);
Sent at Tue, 21 Jul 1998 13:09:43 +0000;
About the security in NSCom 4.5
I'm begining to wonder if this Beta release is to test the browser, or to test the people who use this browser. :)
Keep up the great work. I've modeled a couple of projects off of the layout and work flow concepts at Scripting News, though I've implemented them in Perl using the Web for the UI.
It's under Preferences | Navigator | Smart Browsing. The problem is, the default is enabled...
From: firstname.lastname@example.org (Dave Fassett);
Sent at Tue, 21 Jul 1998 12:21:46 -0700;
Smart browsing can be turned off
Just read some of the article about the Front Page 2000 product, what I think would be cool would be to use Frontier as the storage and scripting engine for all of those xml documents.
From: email@example.com (Joshua Lucas);
Sent at Tue, 21 Jul 1998 08:58:16 -0500;
Front Page 2000
If MS supports xml-rpc then I think that could be one of the coolest things in terms of workflow and site storage, maintenence.
Admittedly I haven't had time to thoroughly review your site (I will tomorrow), but I wanted to immediately say thanks for the mention of keyword.com, and that I couldn't agree with you more on what I've read about the Netscape issue.
From: firstname.lastname@example.org (keyword.com);
Sent at Tue, 21 Jul 1998 00:27:46 -0700;
The Lynx browser has had this functionality for a long time; they call it a "Jump" table. Depending on how the site administrator sets it up, the list can be modifiable or not by the users.
From: email@example.com (James Stansell);
Sent at Mon, 20 Jul 1998 21:21:49 -0500;
Re:Netscape and Generic Names
No, I don't believe Lynx "jumps" have yet gone to a central "portal" server for the URL information. But the point is that it would be a relatively minor change to do so. I imagine some people are already using a local jump server just by putting the file on a network file system.
I agree with you that Netscape should continue to support the old behavior. I don't like your idea of splitting the URL field in two though. But the behavior could be controlled by a preference selection and/or new button. It would be good if a user could choose the other behavior than the default behavior without having to change the default behavior (i.e. pull up and navigate the preferences screens.)
BTW, the Internic doesn't maintain any "keyword" database. The idea of taking an unknown hostname and performing multiple DNS "guesses" to arrive at a working URL was truly bizarre. So is calling it a geek feature. I think most geeks would call it a marketing feature!
The truth is, both the DNS guessing feature and the keyword lookup feature are grotesque replacements for the feature that is really needed: a distributed directory service. X.500 and LDAP provide the framework for the directory, but until it starts to be populated, we'll keep seeing weird "features" that abuse common sense.
I'm glad to see someone is starting to make some practical use of XML in real-world situations. I can't help objecting, though, to your comparison of the XML-RPC "standard" and CORBA in the XML-RPC for Geeks DaveNet piece.
From: firstname.lastname@example.org (Yates, Sam);
Sent at Mon, 20 Jul 1998 16:47:52 -0700;
XML-RPC vs CORBA
XML-RPC is not really in the same category as CORBA. It may have facilities similar to the IIOP standard used in CORBA 2.0 for communicating between systems, but there are many facets to the CORBA standard that cannot hope to be covered by a standard with a 1-page spec.
Let's try to compare apples to apples, and keep up the good work. It sounds like fun
I think it's sad, but telling, when you "change your mind" about the usefulness of a feature you lobbied for only 2 months ago because now that you see it implemented, it has the potential of taking away traffic from your site. How selfish and short-sighted! I expected more from you...
From: email@example.com (John Grohol PsyD);
Sent at Mon, 20 Jul 1998 19:11:04 -0400;
Sad... On Netscape 4.5
The Jini specs are out; it's basically the idea of Telescript and Magic Cap but written in Java.
From: firstname.lastname@example.org (Wesley Felter);
Sent at Mon, 20 Jul 1998 17:47:59 -0500;
Jini specs available
Jini offers both agents and RPC-style interaction; I'm skeptical that moving code around will be more efficient than RPC, but time will tell.
Sun says that their implementation will be Open Source (unlike their Java VM and class libraries), but they also appear to have patents pending and their Web site mentions the phrase "platform royalties". IMO, royalties aren't a good idea when building a "platform", especially one which is to be ubiquitous. It will be interesting to see if Sun tries to have it both ways.
I too was extremely upset with Netscape's recent actions. We'll be launching a site at hosting.com soon, and Netscape will only cause confusion and steal traffic.
From: email@example.com (Host Find);
Sent at Mon, 20 Jul 1998 18:43:15 -0400;
Victim to Netscape's Keyword Feature
I've coined a term for this- Keyword Exclusion. Let's hope Netscape returns one-word keywords back to where they deserve.
Netscape could do some interesting things with their keyword-based lookup system, some things that would be an improvement over the existing shortcut of appending "www" and "com" to a URL 'stub' that one types into the location box.
From: firstname.lastname@example.org (Adam Rice);
Sent at Mon, 20 Jul 1998 16:55:16 -0500;
netscape and generic domain names
I have the "crossroads.net" vanity domain. Someone else has the "crossroads.com" domain. Using either the old shortcut or the new keyword lookup, Netscape will take you to the crossroads.com site, not mine. I don't really mind, but my site is an equally valid "response" to the "crossroads" keyword.
Netscape could have the browser pop up a second window showing some likely candidate URLs that are known to be valid. This window might even include the meta descriptions from the sites' front pages. Perhaps Netscape would automatically go to the ".com" site as before. Perhaps it could monitor user clicks, and let people vote with their mice, so that, say, if a ".net" site got a lot of clicks from the secondary window, that would become the first candidate to show up in the main window.
As to the "auto-complete" behavior vs the "keyword-lookup" behavior, my own preference would be for single words to be auto-completed (either as before, or better yet, using a series of user-defined rules) If the string in the Location field includes a space or a question mark, then it should be treated as a keyword-lookup. A pair of buttons next to the field also would be available for more mouse-oriented people to select these behaviors.
Dave: there might be an easy way out of the forest on the keyword issue; convince Netscape to have their browser try to locate http://www.wherever.com/ BEFORE submitting "wherever" as a keyword to their new feature.
From: PReiber@PrismSolutions.com (PReiber);
Sent at Mon, 20 Jul 1998 13:55:22 -0700;
Re:Netscape and Generic Names
That way the names on the net take precedence over the keywords, as they should. If they don't buy that, maybe there could be a keywords v. domainnames toggle in a preferences sheet?
To the best of my understanding, one of the principal features of a browser is that it has a URL entry field; if they don't resolve user input into that field as a URL they're just plain breaking the rules of being a browser.
Jonathan Hendry expressed surprise that no one has used stock symbols in place of URLs. The interesting thing is that while URLs use domain names to locate a server, the model that the IETF has developed makes URLs just one form of URI. Other forms of URIs can also be defined, including a general class of URNs (Uniform Resource Names). A URN is a long lived identifier that doesn't point to a specific server, but rather to a "thing", no matter where it happens to be on the network. Many existing namespaces can be mapped to URNs- Some experiments I have seen include using ISBN numbers, domain name registry codes (people who register domain names are assigned codes like "DW827" by Network Solutions), RFC #, or even stock symbol.
From: email@example.com (Alex Hopmann (Exchange));
Sent at Mon, 20 Jul 1998 12:36:17 -0700;
If they actually get deployed URNs could be an important part of the network architecture since they provide a Standards based framework for creating new namespaces and providing resolution of those names that can scale to the worldwide Internet. A URN can get resolved into a resource description, and/or one or more URLs that point to the actual resource, but this process can include enough information so that your computer could actually find the server that is closest to you on the Internet out of many servers that mirror the given resource.
Considering the tone and scope of the suggestions for how Netscape, or any browser or search engine vendor, could implement a ``better'' keyword feature really highlights how little the InterNIC or the Internet Domain Name System has to do with the issue.
From: firstname.lastname@example.org (Ken MacLeod);
Sent at 19 Jul 1998 23:04:50 -0500;
Netscape and Generic Names
Domain names are simply one step removed from numeric host addresses, just as street addresses are one step removed from postal/zip codes. The upcoming addition of several new top-level domains was already sounding the death knell for default-to-com that (just about only) Navigator and Internet Explorer support.
I've been noodling about this issue since I noted it and your original DaveNet piece pointed out the issue vis a vis scripting.com. I own gui.com. It's based on the name of my old company, which I no longer maintain. I always thought it was weird that if I typed "cnet" into the location window, Netscape would ultimately take me to the right place but if I typed "news" it would open my News reader. Your advice in today's Scripting news wrt making this a user-selectable option is right on, I think, but experience shows most users will still leave the default unchanged so it's not a perfect solution.
From: email@example.com (Dan Shafer);
Sent at Mon, 20 Jul 1998 10:58:56 -0700;
Re:Netscape and Generic Names
Here's how I'd think about it. Let the user modify specific links on the fly. User types "scripting" and doesn't end up where he wants. So he clicks on a button that says, "No, that's not what _I_ mean by scripting. Here's where I want to go when I type scripting" and then goes to the site that he wants. And it changes. This would also allow users to create their own keywords. There are, of course, lots of issues around this (like where does the repository live) but it may ultimately be the only way to deal effectively with keywords other than simply killing the idea.
I just read your XML-RPC for Geeks. I kept looking around for the specs. I then realized that I had just read the spec. The most important sentence on the page is this:
From: firstname.lastname@example.org (Daniel K. Allen (Visual C++));
Sent at Mon, 20 Jul 1998 10:56:31 -0700;
Re:XML-RPC for Geeks
This page provides all the information that an implementor needs.
That's very cool! Simple. I like it.
Remember, Netscape is the open source browser.
From: email@example.com (Paul Hoffman);
Sent at Mon, 20 Jul 1998 10:54:48 -0700;
Re:Netscape and Generic Names
Not yet, it's not. Communicator 4.5 is not open source, and we're talking at least 9 months before there is a Netscape-branded open source browser even by their own estimates.
This, of course, is part of your problem. If it was really open source, you could just replace their database (or at least one line in it...).
Regarding your DaveNet piece, here's the functionality I'd actually like to see in a browser. Suppose I type "Ford" into the Location bar or in some other "search" bar on the browser. Before I go anywhere, I want the browser to return a pulldown menu of:
From: firstname.lastname@example.org (Paul Blair);
Sent at Mon, 20 Jul 1998 10:54:08 -0700;
Re:Netscape and Generic Names
Ford Motor Company
Ford County, Kansas
Other (maybe leading to a "second-tier" list)
When I'm searching on "Ford," one thing we can be pretty sure about is 99+% of the time I don't want results that combine all of these together. This would be a good way of pre-filtering a query without forcing users to enter all that weird boolean syntax they're afraid of.
Of course, the pulldown list couldn't list everything, and some people would still be unhappy. But from the user's point of view it would be far more convenient than the functionality Netscape is implementing now.
In this discussion I've seen you make comments like the question about whether anyone should own generic names such as scripting.com. But scripting.com is *NOT* a generic name, it is your name. On the other hand, if Netscape's keywords feature directs people away from your site when they type "scripting" into the URL box, they aren't taking traffic away from you because "scripting" by itself *IS* a generic name. If people want to go to scripting.com that is different from looking up "scripting" as a keyword.
From: email@example.com (Michael Dillon);
Sent at Mon, 20 Jul 1998 10:31:34 -0700 (PDT);
Netscape's "geek" feature
The whole point of the domain name system registries is that people can register a unique name as an address. The system is geared so that it is impossible to register a generic name however you can register a name that has generic components such as scripting.org. If some people feel that this is giving you unfair control over the word "scripting" then the real solution is to create more top level domains so that more people can register simple names using any given generic word. For instance, if the DNS allowed names like apple.orchard, apple.records, apple.pie and apple.com then we would all enjoy the same flexibility that we now enjoy with corporate names where we combine two words, one semi-unique and one generic. For instance, Userland Software, Memra Communications, Washington Post, Kepler's Books, Stanford University and so on.
I'm surprised nobody has started a service that would allow the use of stock-market symbols in place of URL's. AAPL instead of www.apple.com. MSFT instead of www.microsoft.com. T instead of www.att.com
From: firstname.lastname@example.org (Jonathan Hendry);
Sent at Mon, 20 Jul 98 09:34:41 -0500;
Studying the works of great programmers by looking at the code is like studying the works of great bridge builders by walking across the bridge. Asking people to comment on programs that they think are "great" simply based on their outward observations is not a valid way to determine the "greatness" of its inner workings. There is rarely a correlation between the outward appearance of an application and the quality or elegance of its innards. Some of the most beautiful examples of software engineering I've seen are applications that have no user interface and no way to measure the "greatness" other than by the fact that it works at all.
From: email@example.com (Chuck Shotton);
Sent at Mon, 20 Jul 1998 09:21:16 -0500;
To study the works of a "great" work of software, you really have to study the original requirements, design documents, implementation rationale, the trade-offs that were made, and then you can study the actual code. But examining a great work of software simply by observing its outward appearance or a cursory examination of the code itself is a superficial examination at best. You'd learn very little as far as being able to repeat the process yourself. As with the bridge analogy, simply examining the girders, rivets, concrete, and steel is not going to teach you anything about how great bridges get engineered and constructed.
I have looked over your XML-RPC Specification document and found much that I like; however there is one thing bothering me. Namely, why can't a '<methodResponse>' contain more than one value? I would think that the response from a method call might sometimes return an entire object's state. In which case the '<methodResponse>' would need to be as rich as the '<methodCall>' format...
From: firstname.lastname@example.org (Jack Bell);
Sent at Sun, 19 Jul 1998 17:51:24 -0700;
Re:XML-RPC for Geeks
Perhaps this is because you are linking RPC's closely with classical method calls which can only return one value. But all true (and many bogus) object oriented languages allow that one value to be an object itself -- with all the concurrent complexity. By limiting the response to a single value you are saying, in effect, that all objects must reside on the server. And that doesn't meet my personal vision of a distributed computing environment.
If you need to return more than one value, package it up into a <struct> and return that. <struct>s can be arbitrarily complex. This should be on the FAQ page that doesn't exist yet. DW
Thank you so much for posting the XML-RPC spec at a geek level. I have been following this topic on Scripting News since it started but I couldn't understand your excitement. After all, XML is a *document* formatting language, right? Even your "newbie" guide didn't do much for me. I guess I couldn't get past my mental image of what XML "really is".
From: email@example.com (Kurt Granroth);
Sent at Sun, 19 Jul 1998 10:51:06 -0400;
The spec finally did it for me! This is totally cool! The KDE Project (which I am working on) has been struggling with using CORBA as the standard RPC system. In comparison, XML-RPC is *much* simpler while still retaining the power. All we would need is a mini-web server in each app (trivial if done in a standard lib) and our message passing problems are solved.
Wow, this is exciting!
|This page was last built on Tuesday, July 21, 1998 at 1:12:42 PM, with Frontier version 5.1.2. Mail to: firstname.lastname@example.org. © copyright 1997-98 UserLand Software.|