2008-01-22

ExternalID is not related to ECO_TYPE

I just thought I'd blog about this because it seems to be causing some confusion.

ECO_TYPE
Use:
In integer used in the database to identify the type of object a table row represents.

Obtained from:
The ECO_TYPE table using the name of the class.

Life span:
Permanent. The name of the class in ECO_TYPE changes when you rename your class, but the integer value never changes.


ExternalId
Use:
A way of identifying a class instance across EcoSpace instances. Kind of like a string version of a pointer to an object.

Obtained from:
Index of the class in TypeSystem.AllClasses + "!" + the ECO_ID of the instance.

Life span:
Although the ECO_ID never changes the class index might when you add or remove classes within the model.


I just want to point out that ExternalID and ECO_TYPE are unrelated. When you change your model it does not mean that the ECO_TYPE in your DB must change. Here is an example

01: You create a model with 2 classes in it. The first class is the root, the next is a subclass
ClassA <---- ClassC

02: You generate the DB for the first time. This creates the ECO_TYPE table data with the following values

ECO_TYPE CLASSNAME
1 ClassA
2 ClassC

03: You create an instance of ClassA and ClassB then look in the ClassA table in the DB, you will see something like this

ECO_ID ECO_TYPE
1 1
2 2


04: Retrieving the ExternalID for each class will return 1!1 and 2!2. This looks like the ECO_TYPE but it isn't!

05: You now change the model like so
ClassA <---- ClassB <---- ClassC

06: You evolve your DB.

The result of this is the following

A: A new entry in ECO_TYPE to enable ECO to see when a class is an instance of ClassB

ECO_TYPE CLASSNAME
1 ClassA
2 ClassC
3 ClassB

Note how the types are unchanged. There is no specific order to this list at all. It's a first-come-first-served kind of arrangement.

B: Your ExternalID's now change. The class index returned for each class is
ClassA = 1
ClassB = 2
ClassC = 3

Note that these have no relation at all to the ECO_TYPE. Just because they are both integers means nothing :-)

I hope this clears it up. Remodeling does not mean all your DB type data needs to be updated!


Pete

2008-01-20

Why I don't use ExternalID for URLs

An External ID consists of two parts, !

ClassID isn't really required for ECO controlled DB structures because there is always a root table containing the ObjectID + ClassID, but when you use ECO to map to an existing DB structure there needs to be a way to know that object 1234 needs to be fetched from the PERSON table.

So, now that we know why the ExternalID consists of two parts on to why I don't use it (much). If I am writing a website with ECO and I use the ExternalID for my URL like so

www.mysite.com/showarticle.aspx?id=23!1234

The number 23 is determined by looking for the class's index in the list of all classes in the model. This is the problem! If you change your model by adding a new class then your index may change from 23 to 24. Not a big problem for your application, but persistent links to that URL from other sites such as Google will no longer work.

An ExternalID is just like an object's pointer address. When you restart your application that address is no longer valid. ExternalID lives longer than an application restart, but doesn't always live past an application remodeling as explained above.

My tips are as follows:
  1. For passing object instances around in code, just past the object instance itself.
  2. For passing a reference to an object to another EcoSpace instance in the same application use ExternalID.
  3. For storing away a reference for an unknown period of time (config file settings, URLs) add a unique attribute to your class such as an AutoInc and retrieve it using that value in an OCL evaluation.
URLs are really the only example I can think of where an ExternalID is persisted externally and expected to live longer than the life of the current model. In this case I think the following URL is much more pleasing to the eye anyway

www.mysite.com/showarticle.aspx?id=1234

2008-01-17

RSA signed streams

I'm writing an app that needs me to be able to write to a file and make sure that nobody changes it. The idea that I had was to create 2 classes:

  1. WritableSignedStream:

    Accepts a targetStream parameter + a privateKeyXml string. I add some space at the beginning of the target stream to reserve it for header info. You then write to the stream as normal (it reports a smaller Length than that of the target stream so your app is unaware of the header). When you close the stream it writes the public key to the stream + an RSA encrypted hash of the data part of the file.
  2. ReadableSignedStream:

    Accepts a sourceStream parameter in the constructor. Just before you read for the first time it will compute an MD5 hash of the data part of the file, and then compare it with the signed hash in the file (which I first decrypt using the stored public key).

These classes therefore provide two functions:

  • You can sign a stream, the app can check the stream was signed by you by inspecting the public key within it.
  • You can additionally pass a publicKeyXml to the ReadableSignedStream constructor and it will use that instead of the public key within the outer stream, this allows you to ensure nobody has altered your file in any way.

Anyway, to the point. I could always use RsaCryptoServiceProvider.Encrypt() to encrypt the MD5 hash, but whenever I tried to DeCrypt() I would get an exception telling me that my public key was a "Bad key".

It annoyed me for a while, but then I realised something pretty obvious really. The way RSA works is that you are supposed to Encrypt using the Public key and DeCrypt using the Private key. I was using it the wrong way around. The solution was simple....

(Signing)
RSACryptoServiceProvider RsaProvider;
RsaProvider = new RSACryptoServiceProvider();
RsaProvider.FromXmlString(privateKeyXml);

MD5 md5 = new MD5CryptoServiceProvider();
byte[] md5HashBuffer = md5.ComputeHash(dataStream);
byte[] signedMD5HashBuffer = GetSignedMd5Hash(md5HashBuffer);


private byte[] GetSignedMd5Hash(byte[] hashBuffer)
{
RSAPKCS1SignatureFormatter rsaFormatter = new RSAPKCS1SignatureFormatter();
rsaFormatter.SetKey(RsaProvider);
rsaFormatter.SetHashAlgorithm("MD5");
return rsaFormatter.CreateSignature(hashBuffer);
}


(Verifying)
MD5 md5 = new MD5CryptoServiceProvider();
byte[] currentMD5HashBufferForFile = md5.ComputeHash(dataStream);
RSACryptoServiceProvider rsaProvider = new RSACryptoServiceProvider();
rsaProvider.FromXmlString(PublicKeyXml);


RSAPKCS1SignatureDeformatter rsaDeFormatter = new RSAPKCS1SignatureDeformatter();
rsaDeFormatter.SetKey(rsaProvider);
rsaDeFormatter.SetHashAlgorithm("MD5");
if (!rsaDeFormatter.VerifySignature(currentMD5HashBufferForFile, FileEncryptedMd5Hash))
throw new Exception("Someone has messed with the file");

I hope someone finds this useful :-)

No symbols loaded

This has been driving me mad for hours now! Whenever I run my PocketPC compact framework app I cannot debug it! None of the breakpoints will stop, each breakpoint just shows as an empty circle instead of a solid one.

So, what was the solution? I tried deleting all PDB files on my hard disk but that didn't do it. In the end manually deleting all of the files previously deployed to my PPC did the trick. Maybe VS couldn't overwrite them or something? No idea, but at least it works now :-)

2008-01-15

Geek quotes

This one is probably my favourite:
"There are 10 types of people in the world, those who understand binary and those who don't".

Here is one I came up with myself some years ago and used to use in my newsgroup signature:
"Blessed are the geek, for they shall public class GeekEarth : Earth".

2007-09-29

Derived properties that you don't want cached

I'm rewriting a Win32 application using ECO. One of the things this application allows the user to do is to specify their preferred unit of measurement for height, distance, depth, weight, and speed.

To achieve this I have an ECO class ApplicationSettings which is a singleton which identifies the units to use. I then have 2 properties for each value....

InternalLandingDistance
DisplayLandingDistance

The idea is that I should never expose Internal* properties in the GUI but display their corresponding Display* property instead. The Display* property will read/write the Internal* property and perform some kind of conversion on it.

If I always store my values using the unit with the highest accuracy (Distance = feet, Depth = millimetres, Weight = Pounds, Speed = KPH) then I just need to convert them by the correct factor when reading/writing from the Display* properties.

This isn't rocket science, obviously I can just implement these as reverse derived attributes right? The problem with this is that I am using a new EcoSpace per form, so if someone changes the ApplicationSettings.DistanceMeasurement in one EcoSpace my derived member has no way of knowing unless I want to put in synchronisation.

Well, my app can work stand-alone or on a network. I don't want to over complicate things, so what I decided I needed was a reverse derived property with no subscriptions + caching, but that just isn't possible.

Well, actually, it is. It's just so obvious! Instead of marking my members Derived I marked them as Transient and HasUserCode. I implement the code like so.....


public float DisplayLandingDistance
{
get
{
return ConvertDistanceToDisplay(InternalLandingDistance);
#if NeverDoThis
{EcoModeler generated code here}
#endif
}
set
{
InternalLandingDistance = ConvertDistanceToInternal(value);
#if NeverDoThis
{EcoModeler generated code here}
#endif
}
}



Now my values will be calculated every time they are read instead of being cached. The "#if NeverDoThis" bit is there so that EcoModeler will always put its auto-generated accessor code within a region that never executes. I could just put a "return" above it, but I don't like those "Unreachable code detected" warnings!

2007-09-10

Mr thicko thicky thickpants

Well, I feel stupid. Worse than that, I generally feel very thick. Especially yesterday!

So, what happened yesterday? I went to Bletchley Park. What a brilliant place! Walking around what was for approximately 50 years one of the biggest secrets of WWII was really quite strange. The stuff these people did was simply amazing!

One guy managed to figure out the inner workings of a machine none of them had ever seen just by looking at two *almost* identical messages enciphered using the same machine configuration. The second message was 5 characters shorter than the first because the operator repeated the message but decided to abbreviate some of the words to save some time. This was all they needed. Not only was the functionality of the machine divined but what seems to be the world's first (semi) programmable computer was then designed + built to decipher messages received by wireless and then transcribed to punch holes. This thing has a optical reader that works at 5,000 chars per second.

The mathematics involved in working this sort of stuff out must have been quite brain numbing. I left feeling quite depressed with myself but it was still a great experience. Standing next to a Bombe machine (cracked enigma codes) was brilliant!