We discussed about the storage and retrieval of hierarchical runtime information in general. In the [url=http://www.asg-platform.org]ASG[/url] use case, we needed to store service-related monitoring parameters (like last operation call duration) for later retrieval. Nearly all developers choose a database as natural solution – but why ?

If you look on the history of standardized information models, you see a long tradition of hierarchical directory mechanisms. The traditional OSI concept of a [url=http://de.wikipedia.org/wiki/X.500]X.500[/url] ‘Directory’ (with its according meta-model, the ‘Directory Schema’) evolved to things like LDAPv3 (RFC 2252/2256), NIS (RFC 2307), CORBA naming service and Active Directory. There are multiple implementations, especially for LDAP, which support a high-performance querying and storage of objects and attribute values in the data hierarchy, usually with one round trip (“Find all users which have the following permission”).

Further investigations on the topic pointed out that everything depends on your type of query. Large complicated queries of highly interrelated data are great for a database query optimization technology. Simple queries for specific objects and values (with filtering) might be better handled with some LDAP-based technology. I don’t know too much about the performance of current database technology, but as a student I learned that database updates are comparatively expensive. This might not hold for LDAP. Again, everything depends on your data access patterns …

At university, we have an increasing problem with literature references to Wikipedia and Google search results. Many students rely their work only on Internet sources, bypassing the classical way of scientific information gathering. If it’s not in Google, it is not relevant.

For me it is funny to see that those people do not understand the problem of information quality management, especially in the Wikipedia case. We encourage the people to rely their literature references on non-URL citations, which is quite simple with things like CiteSeer and the ACM Digital Library. Internally, my professor gave out a quality ranking for information sources:

[list]
[*]Scientific journals
[*]Books
[*]Conference papers
[*]Technical magazines
[*]… (a lot of space)
[*]Wikipedia
[*]Web pages
[/list]

We did an according seminar about fault-tolerant systems, only based on papers which where at least 5 years old. For our Bachelor students, it was a totally new experience, but very exciting. The following (sorry, german) article gives a great explanation of the problem of web information sources:

http://www.heise.de/tp/r4/artikel/20/20982/1.html

We recently researched the communication of different .NET AppDomains (for Java guys: isolates) over Remoting channels. Here is a summary of the results from Martin:

If one application domain is creating another one, then a [i]System.Runtime.Remoting.Channels.CrossAppDomainChannel[/i] is created, with the URI name [i]“XAPPDMN_URI”[/i]. It is based on a [i]MemoryChannel[/i] for serialization. The funny thing is (again) a slight hack for performance optimization. The [url=http://systemobject.com/Object/?o=System.Runtime.Remoting.Messaging.MessageSmuggler][i]System.Runtime.Remoting.Messaging.MessageSmuggler[/i][/url] checks whether Marshalling is really needed for the AppDomain communication. Primitive types, strings and string arrays are simply copied, all other data structures are oficially serialized. The current thread is then moved to the new AppDomain ([i]System.Threading.Thread.EnterContextInternal[/i]) and unpacks the serialized data and all smuggled objects. The [i]ReturnToContext[/i] method moves the thread back to the original AppDomain.

Definition of a distributed system: You know you have one when the crash of a computer you’ve never heard of stops you from getting any work done.
- Leslie Lamport