Customer Reviews

COM IDL and Interface Design

5 star:		(6)
4 star:		(5)
3 star:		(0)
2 star:		(2)
1 star:		(1)

 

 

I am a professional programmer for Microsoft with over 20 years experience. I have been writing COM code for nearly 2 years. This is an excellent book, for the professional. This book delivers what it promises. It's not about C++; it is about IDL and COM.

It might have been more accurate to title this book "Interface Design and COM IDL". The first half of the book focuses only on IDL, a subject that has not been systematically treated in any other COM book. It contains very clear explanations of COM remoting and marshaling and the use of IDL within the remoting infrastructure. The diagrams are excellent. The second half contains a detailed code example that shows both client and server bindings. I wish I'd had this book when I was writing my first VB and HTML plug -ins. It would have saved me a lot of time figuring stuff out, especially the use of alternate identity to implement "property get as QI". The server chapter is a gold-mine of code showing the application of advanced COM techniques. The re-usable ATL code for split / alternate identity alone is worth the price of the book.

Once again Wrox delivers value.

I have been a COM programmer for Texas Instruments for about 2 years. I taught myself how to program in COM by reading various COM books that gave excellent info about COM programming but, very little info about the cryptic looking IDL file and how to properly design interfaces. Dr. Al Major exposes how interfaces should be designed with clarity and reason.

I bought this book having grasped the basics of COM and IDL and expecting it to help me improve my designs and better understand the issues in how to develop a "good" COM interface.

I found to be mostly disappointing from a design standpoint. Many issues which six months later I had learned were really very important (e.g. the vital question of whether to support IDispatch or not and the huge performance penalty in MFC's default use of this interface) were not covered. There are many trade-offs in COM and I found little discussion of them here.

As a reference it has potential, since it is thorough and deep, but as other reviews have mentioned, it's not as easy to use as a reference as it perhaps should be.

In the end, it's been sitting on my shelf largely untouched as other ATL and COM books become dog-eared with use. So from me, it's thumbs down.

Some of the more esoteric stuff in the second half of the book wasn't as useful to me since I pretty much develop standard dual-interface components, but the first half covering IDL and marshalling was extremely good.

CI&ID shows the process of going from the requirements of a system to the interfaces it needs, then using this to implement the COM servers and applications the system will need.

The technique starts with developing a "Specification Model" which leads to a definition of the interfaces that are needed. This is different than developing a class hierarchy based on the requirements of a system and CI&ID shows where and why it is different. The class hierarchy is considered an implementation detail and not part of the design. It also shows how the Specification Model fits into an iterative design process and some to the factors that have to be considered when using it. The Specification Model and COM seem to be made for each other.

Although CI&ID includes a sample application most of the book is devoted to the Specification Model and how all the various COM development idioms fit into it. In fact it is a pretty good compilation of COM coding idioms and their tradeoff on its own. It also goes over the parts of MIDL that are used for COM development.

An implicit system design requirement in the book is that an application should not be able to "devine" underlying object model and why this is important and its implications. In effect it says that applications should not use QI (QueryInterface).

Another requirement is that all useful interfaces should be available to all languages including scipting languages. It explains using Alternate Identities to solve this problem and how to implement them, and includes some enhancements to ATL to do this. The fundamentals of this technique have been discussed on the COM mailing lists, but CI&ID has a fairly complete implementation technique.

It is an information dense book that pulls together COM from a design point of view rather than an implementation point of view. It is not an introductory text and you will not "learn" COM from it. However it will show you a very good technique for applying COM to the design of a system. Even if you find the Specification Model not useful or even wrong, it will give some real insights into the nature of COM as a discipline for writing code.

This book is a bottom up approach to COM, which starts from the IDL, but then goes through all the marshalling mechanisms, remote method calls, etc. About Mid-book Dr. Major presents an application design model which starts from the interfaces. The book will prove useful to those who want to design COM classes and interfaces that know how to behave based on whether they are local or remote, in-proc or out of proc and in general are better adapted to their environment.

I have read many COM books. But none of the books provided the ideas of using COM to design a large scale of system. In Dr. Al Major's book, he revealed the techniques of using COM to design a complex systems. If you want to understand COM interface and especially want to know how to use COM to design your system. This is an excellent book for it.

This is definitely the COM IDL reference book. There aren't many books devoted to the subject and most COM books focus on the technology from a specific language perspective not spending any time on IDL itself. Frankly I can't say that it has helped me a great deal with everyday development but it certainly has made my understanding of a few things much clearer and it has reinforced some ideas I had that I was not sure of. Like other reviewers I did not find the other half of the book to be of great value but it was interesting to compare (partial) implementations of the same design in 5 different languages in the same book. My final praise is on the excellent comparison of the various IDL types and how they map to the various languages including a discussion/description of the ubiquitous SAFEARRAY.

The book topic is a rich one, and I ordered this book before it was published. But the author jumps around from topic to topic, and appears to have a limited handle on the material he covers. What you are left with is a lot of headings with two sentences under them. Worse, imagine a programming book with almost no code examples.

Despite these problems, for the first 150 pages or so I thought maybe I was going to get my money's worth. But the last 250 pages or so are a "case study" that presents neither a complete problem or a compete solution. And it does so in 3 languages.

I feel I know little more than I started with. Save your money.