IRIS Devblog

Danny Ayers mentioned that IRIS is available. It’s time for me to say something. Indeed, I owe readers of this blog an explanation for being rather quiet here. IRIS is the desktop user interface to CALO, and, from time to time, CALO comes up for milestone deliveries. We planned to time our release of IRIS along with the most-recent release of CALO, and, therein lies issues with timing. Releases of large, complex software systems occasionally (read: almost always) entail delays as quality assurance programs find new, always surprising, um, features that need immediate attention. Nevertheless, IRIS is now available. And, I can get back to making blog entries.

Let me talk about (read: manage) expectations. IRIS is huge, as downloads go and as desktop applications go. IRIS is only available, in terms of bootable runtime programs, on Windows platforms with recent operating systems. This is not a Windows 95 program. Memory tells me we have run it on 2003 and XP. Source code is available, from which developers on other platforms might be able to coax IRIS to run. At the same time, there are developer docs on this site and in the downloads which will permit others to begin to add features.

I have IRIS running on an ancient 700 mhz laptop with 256 mb memory. I have no expectations it will perform anywhere near as well (whatever that means) as it does on my 2+ ghz laptop with 1 mb memory. A friend reports

Downloaded and installed IRIS on WinNT, 350 MHz, 384 Meg RAM system.
Running IRIS, very long startup… error message several times that
gdiplus.dll not found (this is something that comes with XP?). But it
seems to be running…

Installation startup on my 700 mhz laptop took 12 minutes. There were a few error messages along the way. The program runs anyway; that’s not to say we are ignoring error messages. One error message at a time…

IRIS is, first, a research platform. CALO is the research IRIS supports. The designed-in domain which the IRIS ontology supports is that of an office environment, which implies (it says here) that the platform, as is, should be able to support many different domains, from general office, to research, to financial management, and more. Just don’t expect IRIS to do all that anytime soon.

Let’s talk version numbers. What you download now says “2.3″. Those numbers related to the 5-year CALO project. Said project is in year 3, and, theoretically speaking, at the end of this year’s work, IRIS would become “3.0″ or somesuch. But, we are having discussions at SRI about converting the IRIS version numbers back to something closer to that which the open source community is familiar. In that case, IRIS is not, in any sense of the word, version 1.0 at this time; closer to 0.3 or thereabouts. Perhaps the IRIS build platform will be tweaked soon to reflect that.

Finally, back to expectations. As mentioned above, IRIS boots and, if you watch the console closely, every once in a while, some error message flashes by. My own heuristic: don’t panic. We probably saw it already. IRIS must refresh its user interface after a user gesture (clicking on something) is completed. But, IRIS is always busy in the background doing things. So, screen refresh is sometimes slow. Delete an email message or click on the “Junk” button to send spam off to the promised land will remove the message quickly, but the “unread message” counter won’t get around to noticing the change for a while (that’s just one example). I’d like to also out a favorite nit of mine, one that has to do with the use of Mozilla. Mozilla has an interesting behavior which causes it to, um, spasmodically take over the screen. That’s fine, so long as you aren’t doing something else. But, if you are busy composing an email or otherwise using the desktop, it’s annoying when Mozilla snaps to the forefront. I have talked with people who quit using Mozilla entirely, in favor of Firefox, for that reason. It’s my hope that we will either figure out what is doing that, or that the migration to SpiderMonkey will mean the end of that issue.

Enough for now. Please keep in mind that this website is a work in progress. We will soon have the email lists up and available for use, along with CVS and Jira for issue tracking. As we progress, we will sort out the myriad of community issues and needs, with the aim to promote a strong and vital semantic desktop developers/users community.

Filed by park at 9:18 am under IRIS
No Comments

Now that I’m back from Galway, and survived the Hacker’s Convention in Santa Cruz, California, I think I can speak freely, now, about events which took place during the SemanticDesktop Workshop. There were many presentations, each separately available, all available as one giant download here. For now, I plan to mention two Semantic Desktop platforms that parallel the vision and ideas of IRIS. The first one to talk about is Gnowsis. The second one, Haystack, I’ll talk about in a following post.

Gnowsis is the work of Leo Sauremann, who coauthored an opening paper for the workshop found here. The Gnowsis group are now sponsoring a forthcoming Semantic Desktop Workshop of their own. Read about it here. The event takes place in Berlin, on 9-13 December, 2005 (real soon now!).

What do we know about Gnowsis, looking at it through the lens of an IRIS designer? Firstly, Gnowsis probably predates IRIS, since our first prototype was crafted in late 2003. Secondly, Gnowsis takes a different approach than IRIS to user interface architectures.

IRIS can be seen from the screenshots to be of the single frame architecture, where we mount all applications, from email to calendar to web browsing in one frame. Gnowsis, by comparison, federates individual applications through a control system and browsers which grant views of the growing knowledge base.

IRIS designers are now taking a serious look at such differences, with an eye to a potential redesign of the IRIS interface. I’ll have more to say about that soon.

Filed by park at 11:26 am under IRIS
No Comments

IRIS was announced and demonstrated at the Semantic Desktop Workshop in Galway, Ireland, today. There were about 100 people in attendance, and several great questions were asked. Possibly the most interesting question went something like this: “What is IRIS doing about the fact that some people want to use individual applications?”

Let’s examine two interpretations to this question.

One interpretation is that people simply want to use their favored client, say, Eudora or Outlook for email. In order to do that, and remain within the design considerations that animate the IRIS implementation, someone would have to instrument those applications such that IRIS can be made aware of semantic events such as new email arrived, read, deleted, and so forth. IRIS would need the ability to cause such clients to display emails selected by users during other IRIS events, such as selecting an email related to some other activity, say, a project or task. That’s a tall task, instrumenting individual clients. Budget constraints often dictate what goes into IRIS; in the case of email clients, the Mozilla client was chosen for its usability and similarity to most other popular email clients.

Another interpretation is that people would rather not have all their clients loaded in one unified framework, as is presently the case. They would rather have each client available to move around on the desktop, or to close when not needed. That particular interpretation is presently subject to much discussion within the IRIS team. Watch this space for more developments surrounding the second interpretation.

Filed by park at 10:09 am under IRIS
No Comments

This post will necessarily be short, simple, and mildly historic in nature, to answer the quesiton: What is IRIS?.

IRIS is a semantic desktop application created by researchers at SRI International and at universities around the United States. IRIS has been created to serve as a user interface and knowledge base for the CALO project. During development, we determined that the IRIS project was based originally on a number of open source projects and could easily be released to the open source community as a complete package. IRIS is, at once, a framework for plug-in applications and services, and a large and growing collection of such applications and services. Of particular interest to developers is the machine learning framework included in the project.

Source code to all components of the IRIS project will be made available following the next milestone build. That build will occur once the latest round of refactoring and cleanup of the source tree has been completed; we expect that to occur in mid January, 2006.

In this blog, several authors will contribute thoughts about the IRIS project, its evolution, the community of IRIS developers and users, and open source technologies in general. We expect to cover topics that include user interface, knowledge representation, machine learning, and more.

Filed by park at 2:15 pm under IRIS
No Comments