Hi, and thanks for stopping by. I read your http://www.smartertools.com/blog/archive/2008/11/20/cloud-computing-challenges-benefits-and-the-future.aspx with interest back in November, and will definitely be keeping an eye on what you’re doing at SmarterTools…

Hi, and thanks for stopping by. I read your http://www.smartertools.com/blog/archive/2008/11/20/cloud-computing-challenges-benefits-and-the-future.aspx with interest back in November, and will definitely be keeping an eye on what you’re doing at SmarterTools…

There does seem to be some emerging clarity around the IaaS/ PaaS/ SaaS/ *aaS bundle, as people become more explicit about whether it’s a *computer* they’re after, a *capability* to build something of their own, an *application* or whatever.

Given existing attitudes to corporate data, I do wonder if we’re more likely to see the semantic problems to which you refer being addressed on private Clouds long before they take hold at scale across the open Web…

There does seem to be some emerging clarity around the IaaS/ PaaS/ SaaS/ *aaS bundle, as people become more explicit about whether it’s a *computer* they’re after, a *capability* to build something of their own, an *application* or whatever.

Given existing attitudes to corporate data, I do wonder if we’re more likely to see the semantic problems to which you refer being addressed on private Clouds long before they take hold at scale across the open Web…

At SmarterTools we are developing products for SaaS–strictly defined. Microsoft’s Azure is likely best referred to as PaaS. Conventional co-location Cloud Computing is likely IaaS (Think Amazon Web Services).

The mind reals. Merriam-Webster provides a Dictionary-as-a-Service, but it is no help to us here.

Be well,
Jeffrey J. Hardy
http://www.smartertools.com

Related article on Cloud computing:
http://www.smartertools.com/blog/archive/2008/11/20/cloud-computing-challenges-benefits-and-the-future.aspx

At SmarterTools we are developing products for SaaS–strictly defined. Microsoft’s Azure is likely best referred to as PaaS. Conventional co-location Cloud Computing is likely IaaS (Think Amazon Web Services).

The mind reals. Merriam-Webster provides a Dictionary-as-a-Service, but it is no help to us here.

Be well,
Jeffrey J. Hardy
http://www.smartertools.com

Related article on Cloud computing:
http://www.smartertools.com/blog/archive/2008/11/20/cloud-computing-challenges-benefits-and-the-future.aspx

Seems there are a few issues to sort through with cloud computing and the semantic web.

1 First, what level of capability is being provisioned — infrastructure (IaaS), Platform (Paas), and/or Software (Saas)? There are some definitional differences depending on which providers you talk to.

2 Second, when you speak about Platform as a Services, then one of the major considerations is whether the applications require transactional logic (as in ACID principles used with databases), or whether the processing can be non-transactional (ala Google, Yahoo, MSN, etc. and anyone else who has to read, index, and search against a very large amount of pages and files, very fast.)

3 Third, the convergence of semantic web of data and cloud has already begun, and there are some very interesting examples of linking together of different sources and kinds of information (e.g., database, document, flat file, web page, etc.) . Here the question is one of how far you need to go to make sense of different forms of information — data, text, table, graphic, image, video, audio, sensor feeds, etc. You need a broad repertoire of techniques that are specific to the type of “sign system” your working with. John Sowa says that “conceptual graphs” become the lingua franca through which these different ways of encoding ideas can be interrelated.

4 Fourth, problems of semantic processing at scale become performance issues. The grail for everyone is to reason in “polynomial time”. That is, as you add more instances, or the amount of knowledge representation, or path length of reasoning increases, the reasoning system performance should grow linearly rather than exponentially (e.g, 2 seconds for the first 10, 2 minutes for the next 100, 2 hours for the next 1000. Semantic web reasoning today doesn’t scale. It may work pretty well for RDFs kinds of inferencing, but degrades pretty rapidly when you start working with the additional semantic of OWL. Cloud computing techniques, it turns out, can help a lot to scale semantic inferencing. This is already happening. However, it means changes some software architecture and engineering practices.

5 Fifth, semantic web (description logic) reasoning itself hardly exhausts the topic of knowledge representation and reasoning. There is lots of interesting stuff people deal with that goes beyond description logic, for example: conditional situation awareness, causality, conflict, uncertainty, counterfactuals, reasoning over goals, desires, and beliefs. Questions of guilt or innocence, or life and death, or conflicting values, are about much more than logical truth or falsity of language statements. While reasoning is often divided into four general categories: induction, abduction, deduction, and analogy, it is more accurate to say that there probably as many forms of reasoning as there are forms of knowledge representation. Computers in the coming era will need to handle multiple kinds of knowledge, information and reasoning as well as combine them. From a performance point of view, the goal remains to handle all forms of reasoning in “polynomial time.” Here too, there have been advances which demonstrate the practicality of this, as well as its applicability to cloud computing.

Seems there are a few issues to sort through with cloud computing and the semantic web.

1 First, what level of capability is being provisioned — infrastructure (IaaS), Platform (Paas), and/or Software (Saas)? There are some definitional differences depending on which providers you talk to.

2 Second, when you speak about Platform as a Services, then one of the major considerations is whether the applications require transactional logic (as in ACID principles used with databases), or whether the processing can be non-transactional (ala Google, Yahoo, MSN, etc. and anyone else who has to read, index, and search against a very large amount of pages and files, very fast.)

3 Third, the convergence of semantic web of data and cloud has already begun, and there are some very interesting examples of linking together of different sources and kinds of information (e.g., database, document, flat file, web page, etc.) . Here the question is one of how far you need to go to make sense of different forms of information — data, text, table, graphic, image, video, audio, sensor feeds, etc. You need a broad repertoire of techniques that are specific to the type of “sign system” your working with. John Sowa says that “conceptual graphs” become the lingua franca through which these different ways of encoding ideas can be interrelated.

4 Fourth, problems of semantic processing at scale become performance issues. The grail for everyone is to reason in “polynomial time”. That is, as you add more instances, or the amount of knowledge representation, or path length of reasoning increases, the reasoning system performance should grow linearly rather than exponentially (e.g, 2 seconds for the first 10, 2 minutes for the next 100, 2 hours for the next 1000. Semantic web reasoning today doesn’t scale. It may work pretty well for RDFs kinds of inferencing, but degrades pretty rapidly when you start working with the additional semantic of OWL. Cloud computing techniques, it turns out, can help a lot to scale semantic inferencing. This is already happening. However, it means changes some software architecture and engineering practices.

5 Fifth, semantic web (description logic) reasoning itself hardly exhausts the topic of knowledge representation and reasoning. There is lots of interesting stuff people deal with that goes beyond description logic, for example: conditional situation awareness, causality, conflict, uncertainty, counterfactuals, reasoning over goals, desires, and beliefs. Questions of guilt or innocence, or life and death, or conflicting values, are about much more than logical truth or falsity of language statements. While reasoning is often divided into four general categories: induction, abduction, deduction, and analogy, it is more accurate to say that there probably as many forms of reasoning as there are forms of knowledge representation. Computers in the coming era will need to handle multiple kinds of knowledge, information and reasoning as well as combine them. From a performance point of view, the goal remains to handle all forms of reasoning in “polynomial time.” Here too, there have been advances which demonstrate the practicality of this, as well as its applicability to cloud computing.