Chris Anderson: Software

Radio is gone!

Mon, 17 Feb 2003 00:58:28 GMT

Note This is the last post to my old RSS feed or Radio site.

Wahoo! BlogX is now my primary software. I think this is good enough. You can get to the old homepage here, and all the comments, etc, are still live there - however I won't be adding new entries.

My new RSS feed is here.

I just tried (for the first time) using my WinForms front end across the internet. Queries are a bit slow. Running off local host is almost instantaneous, so I have to imagine that this is due to network issues. I'll probably have to tweak the application to do more automatic caching.

I'll post new source once I get the bugs ironed out from my live dogfooding!

License for BlogX?

Sat, 15 Feb 2003 18:32:26 GMT

Dare asked a good question in the comments below. What is the license for BlogX? I put a copyright at the top of each file, but I didn't think about enough of what I really want here. I guess I'm looking for a license that provides the following:

Free (no charge) use by anyone with or without modification (free)
Anyone using the source or binaries I provide must hold me harmless (no warranties)
Any product that uses the source or binaries can use any license they wish (no viral licensing)
No limitations on usage in acedemic or commercial spaces (for your use anywhere)

Given that it only took me ~15 hours of coding, I can't imagine there is all that much IP in the source, but I don't want this to come around and bite me in the ass. :)

Any suggestions on a good license? GPL won't work because of the viral stuff... MS Shared Source won't work because of the acedemic limitations... other suggestions?

Better HTML

Fri, 14 Feb 2003 15:12:11 GMT

Thanks to Jorge I got my HTML for BlogX better for accessibility... before I was using spans with classes on them for everything... now I'm using H1...H7 for headings...

Also, I switched the calendar navigation from being a viewstate postback to links with a querystring... better bookmark-ability and navigtion expierence... also F5 isn't broken anymore...

Comment view fixes

Fri, 14 Feb 2003 07:39:05 GMT

Fixed a minor issue with comment view (wasn't showing the comment author)... I guess the last things that is preventing me from switching over is some utility for bi-directional synchronization... i want to be able to edit on the client or server and have both get updated... that seems like it's going to be a bit tricky...

Comments for BlogX

Fri, 14 Feb 2003 07:22:02 GMT

Added comment support to the web frontend for BlogX... wahoo!

Xml Serialization

Fri, 14 Feb 2003 06:21:01 GMT

Hmm... I tried to have BlogX save everything as UTC, however for some reason Xml Serialization seems to have no way to serialize as UTC, so I ended up losing 8 hours between serialize and deserialize... anyone have any thoughts here?

If not I'll ask someone at work tomorrow...

BlogX update

Fri, 14 Feb 2003 06:15:37 GMT

Restructured the UI to use a lot more user controls (such as a single control for viewing an entry) and also fixed the CSS for floating the right pane (thanks!).

Also added basic editing (requires write access for the ASP.NET account on the server to the file system). You can see my first entry that I did with the web front end here.

I tried to follow Michael's advice for using WSYSIWIG HTML editing, but I couldn't get the HTML posted back to the server, so the editing is done via raw HTML. Of course I have great editing in the WinForms front end :)

ok... so i'm not a web designer...

Thu, 13 Feb 2003 07:58:47 GMT

... but I've got BlogX look a bit nicer... and it's all done with CSS, no tables! :)

Don't know how it looks in anything but IE...

BlogX

Thu, 13 Feb 2003 05:42:13 GMT

I'm still using the unpopular day-per-file data model, however I've added web support for categories, permalinks, and history browsing.

Here is the latest.

Next steps...

Category view in winforms
Comments in webforms
Web UI that doesn't look like crap :)

Scalable success

Thu, 13 Feb 2003 03:53:56 GMT

I find it humorous that Dare thinks that our APIs are too "RAD". We continually get feedback that our APIs aren't easy enough. I think that having large "cliffs" between APIs (i.e. having a distinct expert vs. RAD API) is the wrong design.

The problem here is that developers are never just a RAD or expert developer. Personally I like to use RAD tools and APIs at the start of a project, then dig into the expert APIs as I need them. The problem is that when you have two completely different APIs (like VB for RAD, and MFC for expert) you can't easily move between the two. If you start with the wrong technology there is a huge cost to try and switch - and more often than not it can mean completely scrapping a project.

Good API design means serving many different masters.

Syndirella likes it!

Wed, 12 Feb 2003 07:04:26 GMT

Syndirella can read my RSS feed without errors... wahoo! In addition I got comment adding working in the winform front end.

Next steps:

Browsing to more than just today in the web viewer
Comments in the web viewer
Permalinks
Category browsing in both viewers

I'm going to try and not make any progress on this for the rest of the week... I need more sleep... and I've got to start working on work projects at night instead of blog software :)

Work life bal... nevermind.

Debugging, what a concept

Wed, 12 Feb 2003 06:53:59 GMT

Brian got me hooked up with the right error messages and I got my BlogX web view working... still fairly limited. For some reason the RSS feed is coming off as text/html, even though I do:

Response.Clear();
Response.ContentType = "text/xml";

In my Page_Load... this works on my local machine and when hitting the machine with localhost at the site... however when we hit it from the machine name it reports text/html... hmm...

BlogX Scalability

Wed, 12 Feb 2003 05:36:13 GMT

rchildress talks about scalability of xml vs. dbms, etc... just as an example, my test data for BlogX was some dummy data with 24 months of data, 28 days per month, 17 entries per day, average of 8 comments per day... overall it was several hundred megs of data. The system was still fast in the winforms front end...

ASP.NET errors

Tue, 11 Feb 2003 06:33:28 GMT

I'm running into lots of ASP.NET errors with my blog stuff... it works great locally, but when I let Radio FTP it up to the server it gets hosed... i'll have to follow up with my admin tomorrow (that's you Brian! <G>)

BlogX minor updates

Tue, 11 Feb 2003 06:26:26 GMT

Still not ready to release my source for my blogging stuff i'm playing with, but I did get a brain dead simple RSS feed working... (again, this is all temporary, so don't subscribe)...

I thought I'd get some feedback from my architecture...

There is one XML file (*.dayentry.xml) per date. There are multiple entries per date. This allows for queries about what days have logs without ever parsing any files. This means the calendar can be populated with a simple file system query. In addition there is an "extra" file (*.dayextra.xml) per date that contains the user editable stuff (comments, trackbacks, etc.). Each entry can have n categories associated with it, and the system auto-generates a categoryCache.xml file when any entry is modified.

One thing that I need to do is have a mapping from the entry id (guid) to a URL. I figure that I could either change my architecture to have a single file per entry (like Radio), or build some cache. I'm trying to figure out the frequency of hits. I'm guessing that 80% of the hits are on the homepage, so I want to optimize around populating the home page entries and calendar. However, it seems that permanent links are often followed... hmm...

The basic layout for a "dayentry" file is:

<DayEntry>
  <Date>GMT date</Date>
<Entries>
    <Entry>
      <Content>...</Content>
      <Created>GMT date/time</Created>
      <Modified>GMT date/time</Modified>
      <UniqueId>GUID</UniqueId>
      <Title>...</Title>
      <Categories>semicolon demlited list</Categories>
    </Entry>
    <Entry>...</Entry>
  </Entries>
</DayEntry>

What OS am I...

Mon, 10 Feb 2003 07:30:48 GMT

Does this mean I have to move to California?

Which OS are You?

As a side note...

Mon, 10 Feb 2003 07:07:24 GMT

Of course, the link from my last posting is completely temporary... so please don't link to it.

Any real blogger...

Mon, 10 Feb 2003 07:05:46 GMT

There was a saying at Microsoft for a while that to be a real team you had to write you own XML parser... luckily that time has passed. However it seems today that if you aren't writing your own blogging tool, then you aren't a real blogger.

I decided to play around this in my free time this weekend... basically I wrote a set of Xml serializable objects to define my OM. Figured out the basic features, and wrote a WinForms browsing and authoring tool. I got basic browsing, editing, comments (well, 50% of commenting), and categories working.

Finally I wrote a super simple web browser experience... no pretty formatting, RSS, links, etc right now. However I did write a converter to move all my Radio feeds into my new stuff.

I have no clue what I'm going to do with this, but it made for a fun weekend. I haven't been writing a lot of code lately :(

Release cycles... the information side

Sun, 09 Feb 2003 19:22:19 GMT

As Dare pointed out in his comment on my previous post, I only talked about the product bits portion of the problem... what about releasing product information early, often, and releasing control.

Lets look at two extremes - the Xml Web Services team who are working on all the WSI/GXA/etc stuff, and the Windows organization.

The Xml Web Services team is publishing their specifications (and prototypes) to the public, and then working with a working group to actually get broad approval for the spec before integrating it into the product. (Note: this is my understanding, I'm not on that team). Design by process greatly slows the development and design process down. However, it has the benefit that they will get great interoperability with other systems, since everyone will have the same lead time on the implementation of the specifications.

The Windows team isn't (to my knowledge) publishing the specs for the next version of Windows. This has the advantage that they can innovate faster, but of course their aren't the same opportunities for third parties to implement the same specifications to get great interoperability (i.e. getting a Linux version of whatever the next new thing is in Windows).

Of course, the Windows team does engage with a small group of partners to get feedback. The question always is how big of a group, and under what NDAs the discussion occurs. Some people think it is silly to have all these NDAs, but it is a way for a team like Windows to share their designs without having to go fully public - yes?

I understand both of these models to some level. The "design in public view" approach is optimized for interoperability and lack of differentiation. Basically the goal is to have a lot of people able to implement the same designs and work together. The only differentiating factor would be non-standard additions, and implementation quality. To my understanding this is the JCP model. I'd bet that a huge problem in the J2EE space is product differentiation. When you get the Sun execs telling everyone to only target "standard" features, how can one product provide value add features?

The "design in private" model is optimized for differentiation, I would argue. Your competitive advantage is your first to market with design innovations. If every feature that Office or Windows added was first publish with complete specs to the public, would those products ever have time to benefit from their innovation?

As with product bit releases, I don't think there is a black and white choice between the design in public vs. design in private models. I think you want to be somewhere in the middle. Your features that are designed for interoperability should obviously be designed in public, and you competitive advantage features should be help back and designed in private to leverage your innovation. I guess in my heart I'm a capitalist and think that people should be able to make money from the design and implementation of software - not just the servicing of software.

I do believe that Microsoft does too much "design in private"... but I also think that Sun does too much "design in public". I think that the optimal solution for Microsoft and it's customers is for them to move closer to the center.

Moderation is the key.

More on release cycles

Sun, 09 Feb 2003 16:09:10 GMT

Dare sent me an email reply, which I won't repost, but I would like to clarify some things that I said.

First, I believe that MSFT does release to infrequently and not early enough in product cycles. The problem is that nothing is black or white. Large products require large release cycles. Back when we first started on the .NET Framework (then called COM+, Project 42, Lightning, etc.) my team was one of the early internal adopters of the technology - and it was rough. No debuggers, lack of basic functionality, constantly dealing with GC and JIT bugs, interop that worked sometimes, etc. Releasing the product that early would have been devestating. The problem is you can only make a first impression once. If you talk to people that worked on a product like Access that had horrible performance in their first release - they did massive performance improvements, but it still took years for the stigma of the bad release to go away.

I know of projects that took internal customers too early and it is taking an enormous toll. Once you release something (in alpha, beta, or release quality) you have to provide some level of support. Even the act of a release takes weeks. You have to drive the team to reach a quality bar, get full test passes, produce redistributable binaries, etc. Doing a source release is even more taxing - you have to make a pass over the code to make sure it is ready for public consumption (you know, like no swearing in the source code! <G>).

Windows is huge. The source base measures in the tens of millions of lines of source code (I think I heard it is over 50 million?). Driving the thousands of people working on Windows to produce a customer ready (even alpha quality!) build takes time. If you ever managed a large software project, you know that not every build, every day, is ready to distribute to people.

Second, I love source code release. Personally I don't believe that the value of any company (Microsoft, IBM, Sun, etc.) is tied up in it's magic source code. I think that the value of the product is in the support, continued evolution, and innovation in the product. But again, nothing is this simple. There are huge costs to releasing source code. Yes, lots of people do it, and they have structured their organization and development processes around this. Yes, Microsoft is planning on making more and more source code available to certain customers (I read something about smart card access over the web? I really don't know what the plan is there).

In reality I think that Dare and I agree - I just think that I'm a bit more pragmatic about the costs and benefits of releases (both binary and source code). I really do want more frequent, early releases, and I would love to see more source code releases. You just have to understand that these are features - which means that other features would have to be cut.

Bob Dole...

Sun, 09 Feb 2003 02:43:22 GMT

Was just thinking about a funny story... used to work with Marc who liked to have fun with his code... He was known to have creative variable names so when we came across code that looked like:

void SomeClass::SomeMethod(...)
{
    BOOL fBobDoleIsGod;
    ...
    if (fBobDoleIsGod)
    {
...

I always laugh when I think of the code review!

.NET and an industry shift

Sun, 09 Feb 2003 02:23:11 GMT

David Bau talks about .NET as a response to an industry shift [from Scoble]

Release cycles

Sun, 09 Feb 2003 02:00:06 GMT

From Dare:

Release Early.

Release Often.

Release Control.

Good advice, but not always practical when you are tackling big problems. I can really understand how you could release minor versions of applications early, often, and give up control. But when you want to tackle huge issues (like producing a new runtime (CLR)) it is hard to do - especially when you have to support every release for 7 years.

Granted, you could argue that by following the OSS model and giving each customer the source you could relieve the support burden from Microsoft. However, you now have to ask every customer to sign up to support all the software they purchase... not something very friendly either.

I don't know the right balance - I think that Microsoft should release more often, etc - however there is a hefty cost for each release... hmm...

Super stars in the blogsphere

Sun, 09 Feb 2003 01:47:50 GMT

From Megnut:

"Clay Shirky's got a new essay examining Shirky: Power Laws, Weblogs, and Inequality. I've only briefly skimed it but it looks very interesting."

Clay's observations about power laws is really interesting. I especially like the implications of power laws on stardom - "Freedom of Choice Makes Stars Inevitable". What I inferred from this was that given that there are so many people that want to entertain (and be entertained) then you naturally get super stars, primarily because people's choices affect each other. Clay uses this to justify the disparity in the blog community, but it can easily be applied to all sorts of arenas.

Borland does .NET

Sun, 09 Feb 2003 01:27:36 GMT

From Chris Sells... SideWinder - the .NET IDE from Borland... looks interesting...

MVC

Sun, 09 Feb 2003 01:22:43 GMT

Of course doug sides against me... But I mostly agree with Steve Tibbett's comment on Luke's post "MVC is something you build on top of the framework, not something you build into it.."

The building blocks of the platform shouldn't require something as high policy as MVC - but a having an application framework like MFC provide a default implementation of MVC is a good thing.

MVC

Sat, 08 Feb 2003 05:43:57 GMT

Luke chimes in on the discussion

Should Microsoft make a blogging tool?

Sat, 08 Feb 2003 05:14:08 GMT

Rob thinks so [updated link]

Microsoft and the community

Sat, 08 Feb 2003 05:01:56 GMT

I'm still digesting this, so I don't have a good response, but I think this is a posting worth reading...

Nikhil's Book

Fri, 07 Feb 2003 06:41:53 GMT

Forgot to mention Nikhil's book: Developing Microsoft ASP.NET Server Controls and Components. For those of you that don't know Nikhil wrote the ASP.NET data controls, was one of the principle dev leads for all of the ASP.NET controls, and is just a super star all around developer.

Joe moves to Boise...

Fri, 07 Feb 2003 06:16:26 GMT

... but doesn't support perma-links... Joe - fix it!

Owning your own data

Fri, 07 Feb 2003 06:14:54 GMT

John points out:

"That said, the challenges involved in building a "scalable API" are considerably greater than "let's not do MVC". One of the things that I hated about the way classic Windows controls operated was the fact that they held a copy of the data that they owned, rather than delegating ownership of that data to some (and almost always more appropriate) application data structure."

I agree - it's all about the balance... while you want to support cases where the developer keeps an optimized data store (like for a ListView, etc.), you want to continue to support the simple case where the dev pushes the data into the control (like text into a TextBox).

An example of this is the difference between the ListBox and DataGrid in WinForms. Even in the case of data binding we still ended up copying the string data into the ListBox because of the underlying control implementations, while the DataGrid never copies the data, just reads it from the bound source when render is required.

RichTextBox brings up an interesting question. Given the rich editing expierence for the RichTextBox, how could you provide a virtualized store? Perhaps a "ITextProvider" interface that could allow reading RTF from some user store? To make this even trickier, lets imagine the virtual store model for an HTML editor... Trident (sorry, codename for MSHTML which is the system component that does all the HTML rendering, parsing, etc, in Internet Explorer) handles the data and display - using the HTML elements as the model. This is needed for a rich editing experience. How could Trident provide the rich editing expierence with a virtualized store? - unless you virtualize just on the HTML text (with that same "ITextProvider" model)...

... sorry, i'm rambling...

Why aren't blogs exposed with WSDL ?

Thu, 06 Feb 2003 07:34:52 GMT

Are there any plans to create a WS-Blog spec that uses WSDL, etc ?

Microsoft Bloggers being Watched?

Thu, 06 Feb 2003 07:33:20 GMT

Mary Jo is keeping a list of all the 'softie bloggers out there... it's a pretty good comprehensive list... Josh is also keeping a pretty good list...

ASP.NET meets WinForms

Thu, 06 Feb 2003 07:26:12 GMT

I'm sure this is ancient news, but since I gave a shout out to SharpDevelop, I figure I'd do the same for Web Matrix... Nikhil started this tool in his spare time while working at the 'Soft, everyone thought it was so great they gave him a team and he shipped it as a free tool on the web... pretty cool stuff!

WindowsForms.com... now with Forums!

Thu, 06 Feb 2003 07:21:50 GMT

WindowsForms.com has added discussion forums for questions, answers, information, and rants... I know the product team is hanging out there so give them a holler! (well, they also hang out at a bunch of other sites, but I digress...)...

Bedlam DL3

Thu, 06 Feb 2003 07:18:32 GMT

I'd like to add a few more email pet peves to my rant...

First, check where you are sending mail. In Outlook you hit Ctrl-K and it resolves the names - make sure they are right.

Two, limit the number of people on a mail thread to the people that care. Don't CC a mailing list of 1000 people

Third, and most importantly, never respond to a large mailing list with "Me too" when someone asks you to be removed from it.

Why can't smart people figure out these basics?

Scalable APIs

Thu, 06 Feb 2003 07:11:38 GMT

I hate MVC. I mentioned this earlier, but I think I need to expand on this a little. It's not that I'm lazy or haven't read Design Patterns. I just believe that simplicity wins. There are times when using abstraction models like MVC actually make your solution simpler. For example, when creating a word processor and you need to have multiple views of your document available at the same time - this is a great time that MVC will save you.

There are times when having design patterns baked into the platform just make it too hard to do simple things. Swing forces you to learn MVC to do simple tasks. I've heard that SWT fixes some of this, but I haven't got to play with that yet. (In truth, I haven't got to play with Swing yet - I've read some books and attended some JavaOne panels on Swing though).

You can always add complexity to systems, you can never remove it. Once you force a design pattern into a system it always shows through. You can provide layers to try and hide it, but it is always here.

This kind of debate comes up all the time at work. We have such a variety of developer customers some people wonder if we will ever be able to produce an API that works for everyone. People talk about producing a "RAD" API and an "Expert" API... personally I find that a bit insulting - I think that experts are exactly the people that have figured out how important it is to use RAD tools when they are appropraite.

I believe instead in an approach of producing scalable APIs. The API should scale from the begining scenario up to the expert. The API should reward people that learn the inner workings of it. In the same way that a user of EMACS can become 10X more efficient by learning all the archaic commands, so should an API provide unlimited headroom (ok, maybe i'm going a little off the deep end here...)

When designing APIs people should think - How will this be exposed in a tool? How will developers discover the features? What features are the 80% case, and how can I make that simpler? How can a developer progress from the simplest scenario into the full features of the API? Good API design is tough. It is a skill.

To you project managers out there: Good API design costs money. Pay up or the developers will stop using your technology - no matter how good it is.

UI Separation

Tue, 04 Feb 2003 03:53:15 GMT

Interesting article... i'm a big believer in practical programming, so I generally move away from complex MVC type stuff… however the notion of a UI programming language sounds kind of interesting… would be interesting in reading more on this… Thanks for the link Dare.

More on good management...

Mon, 03 Feb 2003 05:50:06 GMT

I know that I sound like a broken record, but I continue to be proud to work at Microsoft. This weekend I found out about an employee (who will remain anonymous until this is announced internally) who needed to move to be with his family. At some companies this would mean having to choose the company over family (an easy choice, your job is always replacable). This guy's manager is working out a deal to let him work remotely. I don't know the details of the arrangement, but it makes me happy to know that if I needed something like this that there are people in the company that will make the right thing happen.