bharry's WebLog

Yesterday the Release Candidate for Team Foundation Server was released on MSDN Subscriber downloads. Upgrading my Beta 3 Refresh installation of Team Foundation Server to the Release Candidate went very smooth.

Carl Daniel (code) and Robert Downey (code) each wrote and posted code to show the changesets between...

A somewhat frustrating issue is that the Find Label dialog only allows you to select team projects and not subfolders under a specific team project -- where the labels are in practice actually applied.

At a minimum, it would be nice to have a label history window (if label history isn't displayed in the changeset history window.

Any idea on an ETA for when this may be improved?  After a conversion from SS to TFS, our version control history is practically unusable.  We have about 180 sub projects (drivers) all with labels like  (V2.35.005)  This makes the find tool useless, as there is no way to search for a particular drivers release.  We will work around this issue by running a script to rename all our labels with the driver name prefix, then converting the SS database again.  Still it would have been nice to search at the subfolder level.

If a label specifies the version of each file, how difficult would it be to add a column to the file history which includes a comma delimited list of labels that reference the specific file version?

This would help a lot in spotting the changes that have been made between two labels.

Mark

Any ideas on when MS will take a non-developer approach to this discipline?  

A developer approach isolates the rest of the disciplines in the SDLC and cause significant limitations for an appropriate end-to-end SDLC solution.

I have over 15 years experience and have never seen a developer-based implementation work well.  I was very excited when MS announced they would be moving away for VSS, but was greatly disappointed with the new set of tools as it demonstrates that MS is still way off target for this industry.

-tk70

CM/SCM is a non-developer discipline with its own set of requirements that version control tools need to account for.

VSS got a pass because it was not created as an enterprise solution---as one of its original creators you could agree or disagree with that statement.

TF needs to have solid involvment from the CM/SCM community to ensure that it and other SDLC-based tools are created properly.  It should be developed based on the right set of requirements, not just what will make developers happy.

Take DirectX for example.  MS created it with little input from the game community originally.  Initially, not a very many game programmers liked it.  After MS involved the game community, DX got A LOT better and is now the standard.  Yes, I can make this claim as I am also heavily involved in that industry.

As a big fan of MS, I would like nothing more than to see the company succeed in this market space, but the current path tells me that MS will have a hard time making the desired impact and while MS plays catch up other established companies are getting better.

Just my thoughts and trying to provide valuable feedback.  This is not a bash post, so please don't take it that way.

Thanks for listening....

The CM/SCM discipline should be involved in SDLC from start-to-finish---although it never seems to happen that way :)

While we don't tell BAs, Architects, PMs, etc. how to do their jobs, we do maintain the artifacts that are generated from each discipline and thus, install, configure, integrate (in some cases), and maintain the tools used by those disciplines.

So, it would be nice to have a complete suite of tools that are tightly integrated---a good example is Office.  Each tool can work in standalone, but is also very aware of the other tools in the suite.

Right now, the general impression of the MS offering is a bunch of existing tools trying to work together.  But they were never really built to play together in the way they are offered today.

I would love to see MS offer a suite like IBM Rational's, but with A LOT less overhead.  They have shown how a tightly integrated suite can be successful, but it is way too bloated and requires a dedicated team to support it.  MS has the power and backing to offer the same type of suite, but has the opportunity to make it slick and easy to use---all without having to spend a ton of money on separate hardware architectures to support it.

Now in offering a tightly integrated suite, MS could give CM/SCM the power we always desire---a shell environment with a very powerful command system.  The ability to write scripts across all tools in the suite for better streamlining of workflows/processes.  No need for Python, Perl, etc.  We could leverage the sheer power of .Net and make some very sweet things happen.  No more Crystal Reports, we could leverage Reporting Server's full power.

This is surface level stuff, but hopefully it is a starting point.  I know TF/TFS is going there because I did take the time to really review your blog.  But I feel MS is just reinventing the wheel with little flare to make a company tank their current investment. I have kept IBM Rational in many, many customer sites once I showed what it could do when you know what you're doing.  I'm just waiting for the same opportunity to standup and show what MS is capable of doing.  Let's face it IBM couldn't do it that's why they bought Rational.  But MS has proven that when it listens, it evolves entire industries (and the world).

tk70

Brian that is EXCELLENT NEWS and thanks for the insight!  I have made a note to follow Orcas and Rosario closely.

Please let me know the best place to track these future releases as I would like to get a jump start on them so when they hit the market I am already up to speed on them and can provide instant value to my customers.

tk70

Will you please give me a date when the label information will be included in the file history?

This is a very important feature, esp. when I'm trying to figure out if a particular file was labled correctly!

I found this discussion very interesting and know where tk70 is coming from. One slight correction - MS was not the original creator of what is VSS. It was developed by a company called OneTree and was on the market a few years in the early 90's when MS acquired them.

Also, the very idea of replacing files in a label contradicts good SDLC discipline. I think the best purpose a label serves is a baseline. Mechanisms that can change identified baselines defeat the purpose of having a baseline. I believe even VSS had the ability to extract a file from a label and replace it with another version. I am guessing this ability only existed because many companies use automated build/compile scripts and require the label intact to kick off the process. If there was a problem with any one file, they might just replace the one file in the label and rebuild, rather than make a whole new label with different version number. For work internal to development, this might be OK. But when releasing to a major test bed or production, this is not a good practice.

PingBack from http://websitescripts.247blogging.info/bharrys-weblog-why-tfs-labels-arent-like-sourcesafe-labels/

Ability to place versions of files from different points in time into the label. How to achieve this?

Been using TFS for about 6 months. The learning curve is extremly high and I'm not happy with how the TFS team has left out so many features. Come on guys, get with it. We cannot see labels with our history? Just foolish. This like so many other things should never have been left out.