Because of a discussion session, I suddenly found a critical bug in #SNMP library. Therefore, a new Work Item is added and I am going to work it out before releasing the UnicornHorn Refresh.
I am not sure if there is a lot to be fixed but hope that I can resolve all this weekend. Stay tuned.
BTW, brianb, thank you for reporting this bug.
There are two significant bugs inside UnicornHorn release that are fixed lately. Thus, I plan to release an refresh release this weekend. The version number should be 1.1.
Major changes are,
Even though the MIB browser has been updated lately, it is still an Alpha baby.
Stay tuned.
I know that it is impossible to write a perfect MIB parser in a few months, but I am still satisfied about the simple parser I wrote for #SNMP library in May and June as it successfully parses 70 Net-SNMP bundled MIB documents without an exception.
However, I know there are bugs inside because Net-SNMP bundles nice MIB documents with consistent format. Yes, other MIB documents may cause an exception just because its format is cripple. That’s my fault, because I never knew how ugly a MIB document might be.
During the development of #SNMP MIB Browser (especially the MIB tree feature), I have to enhance the parser from time to time to parse a few MIB documents downloaded from the Internet. Luckily, the changes are small, so I think my original design is not too bad.
And today, I can announce that this parser can parse over 100 different MIB documents without an exception. Stay tuned.
Maybe I should not have designed Manager.GetTable that way, isn’t it? The trouble I got is more. So how to change the original design at this moment? Take a look at source code of Change Set 14707, and you will see my decision.
The Old Design
Once upon a time (in this April), there was no validation in GetTable. But I added a basic case as soon as I noticed the necessity. This old validation required the OID passed to GetTable could be translated to textual form and ended with “Table”. It is such a natural rule that all tables defined in standard MIB documents follow it.
The Problem It Brings
The old validation is too strong because not all OID can be validated like this. For example, if you try to access a table defined inside a private MIB document and probably #SNMP fails to parse this document, you are stuck! Sorry for the inconvenience.
After hearing of this critical bug, I admitted that I made a mistake when closing Issue and a solution must be presented as soon as possible.
The New Implementation
I think a weaker validation is preferred before I fix all possible issues in the parser. Thus, a new rule is added to the validation,
Hope you like this change. Stay tuned.
I believe the table manipulation is one of the hardest things in SNMP world. Therefore, I have tried my best to make it simple for #SNMP users.
Have you tried Manager.GetTable method? In most cases, a simple call to this method provides you the rows and columns. But I want you to pay attention to some points.
Table OID
The OID you pass to this method must be a valid OID for a table. If you pass in an entry OID or other, I cannot guarantee you get what you want.
Sparse Table
Some SNMP agent does not fill in gaps inside tables. In such cases, #SNMP cannot generate a correct output matrix so the result is wrong. I am already aware of this bug but I do not have time to further investigate this issue at this moment.
The Implementation
The current implementation of GetTable is based on GET NEXT operation. Thus, it is not the most efficient but I hope I can tune it later.
Stay tuned.
I have been reading posts about upcoming Delphi 2009 (code name Tiburon) for weeks, but I love this post from Castalia’s author Jacob.
Yes, because Unicode will be everywhere necessary inside Tiburon (compiler, RTL, VCL, and IDE), Jacob have to update Castalia source code somewhere. I don’t know if the GExperts maintainers have a blog but I know that they have to update GExperts for Unicode, too.
Do I need to update Code Beautifier Collection for Tiburon? I have no idea this time. CBC is written in C# and .NET, so it is already Unicode based. Therefore, if there is no significant changes in the OTA for .NET, I hope I can release a build for Tiburon soon after its birth.
Stay tuned.
Finally the 1.0 release is here.
However, soon I found more issues inside the simple MIB parser with some MIB documents and there were a bug report in the discussion board, too. So, there is still a lot ahead.
I hope I can fix the parser at first and then move on to the MIB browser. Stay tuned.
I have been using Matt’s StyleCop Addin for SharpDevelop for almost two weeks and it is wonderful. So from now on, I am going to enhance #SNMP source files to make the code look better.
Luckily #SNMP is not a too large project and I am just getting 10,000 warnings from StyleCop. Hope that I can clean most of them up this weekend and deliver you UnicornHorn by then.
Stay tuned.
It is really hard to master coding, because you are not just simply implementing a feature of your project. Writing solid code, secure code, and beautiful code is not a day dream.
At this moment, I feel like upgrading my coding style, so I find myself resort to Effective C# a lot. But, it is not always easy to find bad code just by my eyes. Then I feel luckily as a C# developer, because there is a lot of cops around me to help.
Microsoft Code Analysis (FxCop)
This was a popular GotDotNet project, and from the old days I have forced myself to fix most violations it report. Generally speaking, FxCop points out best practices you should follow when using and designing elements. I love the name checker a lot, and also the localization helper.
Because Microsoft decides to merge FxCop functions into Visual Studio (for above-Pro editions only), I am afraid that one day I may not be able to get help from FxCop.
The latest version from Microsoft is 1.36
Microsoft Source Analysis (StyleCop)
StyleCop focuses on source files. It helps you to develop a common coding style with others easily.
The latest version from Microsoft is 4.2. By default it only supports non-Express VS, but there is also an addin for SharpDevelop 3 here.
Gendarme implements certain amount of FxCop features while providing new rules. Because it is open source, you can feel free to extend it whenever suitable.
At this moment, 0.2 is released for Windows. You can find it in Google Group for Gendarme.
NDepend provides a powerful query language named Code Query Language for you to query information from MSIL. It is a perfect way to dig deeper under the surface. However, I always find it hard to master because there are so many important features and tricks. Hope some day I can spare a whole week playing with it to investigate all its abilities.
Conclusion
You’d better start new projects with these cops so they can help you from day one. And if you use them to analyze a legacy project, there might be thousands of violations.
I had a Facebook account, but I rarely used it because I found out that I do not like the FB way of communication. I did not touch MySpace or others until I found LinkedIn on Marco Cantu’s blog. Therefore, I keep updating my profile there and hope that some day an opportunity can reach me. It is always hard to find someone who knows my skills well.
By the way, if you have a LinkedIn profile and want some recommendation from me, leave me a message.
Lately I did not update the browser, and of course users met issues with it and posted on the discussion board here. Thank you so much for posting your ideas. They helps me sharp the browser in design.
But before the 1.0 release scheduled this weekend, I think I have no time to update the browser. Thus in my last post, I proposed a 1.5 release to address the browser issues. As a result, when you see 1.0 some day, don’t blame me for the browser quality.
I talked about the browser plan in details, so if you have any suggestions to add, please append your words there.
Stay tuned.
Lately I have been fighting against the heat around me at home because it is so hot here at Shanghai this summer. Therefore, I did not add new stuffs to the browser. However, did you notice the updates in SNMP v2 support inside the library such as exceptions and TRAP v2 support? Sometimes I feel it much easier to add things in a library than an application.
After so many changes underlying, I think now it is time to prepare the 1.0 release of #SNMP. In the current plan, this release still focuses on the library itself (the browser and other samples are only available in source code form). And I’d like to release the first full feature MIB browser in a new release called TwinTower (1.5).
Stay tuned.
For a long time, #SNMP focuses on SNMP manager side. Therefore, when one user asked if he/she could send out TRAP with #SNMP, I was surprised.
And today, I am so glad to see two votes on my poll that my work last night is not meaningless
If you check out the latest source code, you can see a new project named TestSendTrap. This project illustrates how to send a TRAP with #SNMP. I hope it is a good start on #SNMP agent side support.
The development last night went on quite well because TrapMessage class and TrapV1Pdu already provides the skeleton. All I needed to do is copy one constructor from GetRequestMessage and then paste into TrapMessage. After a few modification, a new constructor for TrapMessage was there. This constructor helps you to construct a TrapMessage from all information you want to send (including enterprise, generic, specific, and so on). At last, a Send method was added to TrapMessage, too. You can simply call this method to send the TrapMessage instance out via UDP.
(please notice I renamed TrapMessage to TrapV1Message because there comes a TrapV2Message)
This TestSendTrap project can be tested against TestTrap project and it worked fine last night.
After all, it is not too hard to add agent side support into #SNMP. However, because of the complexity of SNMP agents, there is a lot to be done. Stay tuned.
I did try Pulse a long time ago, but I was not happy with that. However, after Eclipse 3.4 announcement I tried Pulse again, and the new version, is excellent!
I just could believe so many things were changed by the team,
I guess I would no longer download Eclipse without Pulse :-)
(Updated: please notice if you want to use Pulse free, you should read the first comment below carefully to make yourself comfortable :-))
When I wrote this post, I was upset. And I never knew that Microsoft would change, after Bill Gates’ retirement.
Now, Sandcastle is back to CodePlex, WITH SOURCE CODE. And I think I can see Sandcastle inside SharpDevelop 3.0 soon.
To develop a MIB browser was my ultimate goal on #SNMP. However, at this moment, I find that the hardest thing in fact is the MIB compiler. Why? If you notice, a #SNMP MIB Browser is already on its way to hit you.
Day by day, I keep adding new code to make it more powerful. Yesterday, I have completed the MIB tree panel user interface. So for the first time you can see what MIB objects are parsed from the 70 Net-SNMP MIB documents, and see their types (Object, Table, Entry, Column, and others).
A few bugs were found during the development and fixed immediately. Luckily these issues should not affect normal SNMP operations, and they only affects #SNMP MIB Browser.
Another good news is that I am going to publish the source code of this #SNMP MIB Browser under a more friendly license, the BSD license.
Stay tuned.