Auto-Mapping
-
This is pretty amazing, a tool that maps database fields and encrypts the data with access control tokens. The mapping supposedly helps with ETL type operations where fields of data are the same, but might be named differently, such as "first name" and "last name" in one system mapping to "name" in another. There are also encryption mechanisms mentioned that can help protect information and limit access across organizations, not just within a single company.
I tried to read the paper on this system, but honestly, I wasn't interested in spending 2-3 hours to potentially decode it. It's written by some guys with heavier math background than I. Check it out if you are interested, but I'm skeptical.
The whole encryption thing aside, being able to map fields together is a tricky operation and I'm not sure I'd want to have some tool deciding what information is really in some field. This could be difficult to undo if you trusted it too much.
Even if you had someone review the results, that leaves you open to mistakes if the reviewer doesn't spend the time to carefully ensure the mapping is correct. I'm a firm believer that someone does a better job of doing the work than reviewing what someone or something has done for them.
I think it's a good idea to find some way to maintain access controls between companies and that was the more interesting part of this paper, but I'm still skeptical. A large key infrastructure has proven very difficult for most everyone to implement. I can't help but think that this system either doesn't work that well or has some serious security flaws.
Maybe I'm being too cynical here, but the claims about this software seem too good to be true.
Steve Jones
-
OK, I see it now, the value here (to my mind) is as an independent "object broker", a sort of universal EDI. The automatic mapping seems a little bit of a waste of time, I agree I wouldn't trust it either. I mean, if you're going to share data you need to take the trouble to find out what it is and where it's going, if you don't know that then how do you know who's allowed to see it? Makes the whole exercise pointless surely?
So it doesn't solve any of my problems. Mapping fields I can do in SQL, in fact I prefer to do it there, because when you're mapping fields you're also tweaking data. What I don't have a good view on is which fields are mapped, it's kind of hard to manage once you have a few tables involved, if someone offered me an easy way to do this I might look at it, but I think I'd want it to view the mappings I'd already made, otherwise you're down to on the fly code generation, all I have to say on this issue is "Access SQL generator".
Having said that I control both ends of the equation, if I only controlled one I'd use XML or Forrest or somesuch.
So this is the process for sharing information as I see it :
1) Decide what information to share
2) Make sure both parties have the data in the schema they need to share
3) Transfer the data.
Once the data is transferred access is in the control of the DBA/APP at the receiving end. If the sender doesn't want me to have control of the data they're sending then they don't send it.
So what's the point of this again? Seems to me they're trying to make something intelligent enough to do a DBAs job without actually considering that they're only solving 10% of the problem......
-
Take away the encryption and it's called XML !
RegardsRudy KomacsarSenior Database Administrator"Ave Caesar! - Morituri te salutamus."
-
The following excerpt explains how it really is in 99% of the cases.
"the ontology-mapping table may be generated in either a semi-automated way or a totally automated way. In the first case, an ontology matcher (i.e., a human expert trusted by both organization A and B) will be able to access both A’s ontology and B’s ontology in cleartext. He understands the semantics of both ontologies; he will use specific ontology mapping techniques [19] to map; and he will generate certain entries or rules in the mapping table which we will explain shortly."
This bit of wishful thinking, below, suggests there may be cases (I submit the cases are the trivial examples used in beginning technique classes) where the mappings can be automatically generated.
"totally automatic ontology-mapping table generation does not need a trusted human expert, so it achieves more privacy, although sometimes it may not achieve perfect accuracy (of mapping)."
The reason I am not excited by this goes to the nature of semantics in the business environment, where humans attempting human-to-human communication often fail miserably at a shared definition of the business objects. The expectation that a machine can do a better job at detecting and faithfully absorbing subtle business rule nuances borders on superstition.
Viewing 4 posts - 1 through 3 (of 3 total)
You must be logged in to reply to this topic. Login to reply