Slashdot Mirror
A .Net 2.0 Migration Strategy?
An anonymous reader asks: "I work for a large organization where we use .Net 1.1 as our sole development language. We have many frameworks, applications and web sites that are developed in .Net 1.1, and these developments are by no means trivial since they are the result of an IT department of over 300 people and 2 years of development. It is my responsibility to develop a strategy to move to .Net 2.0, this includes the existing applications, new developments, integration, QA, live and development environments. Does any one have any experience in this (preferably at this scale) and can any one recommend any reading material that would help?"
At work we're currently in the midst of migrating our app to .net 2.0. As trivial as it should have been, a *lot* went wrong - a lot of functionality is different, and it's not even documented. It may compile, and it may run, but not for long.
.net 2.0 and enhancing the product.
All I can say is... 1. get it compiling, 2. get it (apparently) working with some shallow QAing, and 3. pass it off to QAers, because they know what to test better than you do. Only after should you even consider utilizing features of
Buckle your ROFL belt, we're in for some LOLs.
I do have experience migrating several smaller .net 1.1 apps to 2.0, but nothing even close to this scale. I will give you as much information from my experiences as possible, and hopefully it'll be of some use to you.
a milyID=7cecd652-fc04-4ef8-a28a-25c5006677d8&Displa yLang=en
The WinForms upgrade wizard worked pretty much flawlessly for each application I converted over, the trickiest part was setting up Source Control, and it was more of a problem with the source control solution we used (SourceGear). Good source control system once it's setup, but a real pain in the ass until then. There were 3 or 4 fairly large projects and a ton of small ones that all converted without a problem.
Not sure if you're using the web stuff or not, but that was where I had the most problems. The Web upgrade wizard is far from perfect. You should at least install the latest version of it from: http://www.microsoft.com/downloads/details.aspx?F
That version has fixed several issues I had, but also seemed to introduce a few new ones (or it was a cascading problem, that resulted in different errors when the originals were fixed). I've had to do a lot of tweaking on many aspx pages to get it to find the Code Behind and there were a couple projects that I had to re-assign all my event bindings with, for some reason, although that was with the old version of the conversion wizard so that is probably fixed.
If SQL is being involed in this upgrade procedure, that was a much harder process to get migrated. Tables, Views and Stored Procedures migrate fine and the Reporting Services reports migrated over.. ok (though VS2005).. but the HUGE hassle is analysis service, if you use that, don't expect it to migrate very easily. DTS stuff was pretty easy to migrate over.
Anywhere, theres my memory dump. Hope that helps in some way.
Wait... and you're saying they got something done? That's not how it's supposed to happen!
(cynic)
Village idiot in some extremely smart villages.
That's what you get for using Microsoft app platform. By the time you will move everything to .Net framework 2.0, they will have 3.0 released and you will have to go through this all over again. When I paid over $2000 to get Visual Studio 2002, little did I know that it only supported .net Framework 1.0. When v1.1 came out, Microsoft said you have to buy Visual Studio 2003. Now that v2.0 is out, Miscosoft says you have to buy Visual Studio 2005. Your VS2002 or 2003 won't take .net framework 2.0 and they don't give you a tool/patch to make it work. Just keep buying and paying $$'s...
One of the major potential benefits of using class frameworks and object-oriented programming is to free code from heavy dependencies on underlying infrastructure through abstraction. And yes, I know it's not mandatory or anything, but the potential is staring you in the face and so it's silly not to take advantage of it. With .NET (and MFC, etc before it), Microsoft can't do too good of a job of this because to do so would take you off the Windows / Office / Servers / Visual Studio upgrade treadmill, from which they derive their income. There are plenty of libraries and frameworks out there that will allow you to build for Win9x (still!), NT/XP/Vista/2Kx/.net, OSX, Linux, etc. that offer exceptional stability between versions.
Yes, Microsoft has some decent tools out there (Visual Studio has come a long way since the first version), but their behavior here has been consistent for the two decades or so that Windows has been out - you will wind up doing some significant porting and testing between versions of their tools, because compatibility between versions is not a priority for them. Selling you upgrades it the priority. Adding value is not a priority - they'll give just enough additional value between versions to make enough people jump, thereby forcing even more people to jump to maintain compatibility with the first set. It's an endless treadmill. If you (or your company) chooses to spend the immense amount of time and money to run on this treadmill, that's perfectly fine with me. I'll just sit over here accomplishing far more per dollar of CapEx and per man-hour.
Help save the critically endangered Blue Iguana
Side-by-side installation hasnt been a problem either. Both frameworks can co-exist, with a few tweaks here and there. The language (C#) has gotten a bit nicer. More shortcuts, faster development, and overall superb IDE support(VS.NET 2005). The deprecated features have been done for a reason, and overall the changes make sense. While the performance is a bit better, I dont know if its enough to make a business case. If I can, I would wait a bit more, till a Vista release; especially if I'm doing WinForms apps.
I see a lot of people saying things along the lines of "thats what you get for sticking with Microsoft". I think this is a somewhat unfair statement.
.Net 2.0 makes 1.1 obsolete. I know that programmers are attracted to all the "new and shiny stuff" like lemmings to a cliff, but if logically speaking, unless there is a true BUSINESS REASON for porting an application, why not just leave it where it is. It is not like Microsoft is suddenly going to pull the plug on 1.1. It'll still work just fine and dandy. In fact, you can run a 2.0 and a 1.1 application pretty much side-by-side, so there really isn't any issues here.
First off, it's not like
Second, it's not like Microsoft is the only one coming out with new revisions every few years. This is pretty much the same strategy employed by most software makers, including the open source folks. Some people have brought up Java as this great and wonderful language. I am not a Java expert but I know that Java has gone through several iterations and during that time, many of their libraries and frameworks have changed. Functions have been deprecated. Libraries and their API's have been rewritten. New features have been added. I am not saying that this is a bad thing, but it is strange that people throw Microsoft under the bus, while giving Sun a pass.
I think the migration path from 1.1 to 2.0 is fairly simple from what I understand. There are various tools and wizards to convert over some of the changed libraries. For the most part, though, the core language hasn't changed. It is only the associated libraries. I say for the most part since there have been certain enhancements to the language, like generics (which I think was also added to Java at some point as well, so again there is a double standard here).
I think if one were going to debate that Microsoft has too much software churn, a better example would be the change from VB6 to VB.Net. In that case, there was a much bigger change which very little migration path, other than interoperability and re-writing code. From my point of view, VB6 was totally toss into the trash can there. In contrast, the path from 1.1 to 2.0 really isn't that big of a deal. Microsoft took a lot of flak for the VB6-to-VB.Net debacle, so maybe they learned their lesson.
------
www.moneybythenumbers.com
One thing I haven't seen discussed here is taking a strategy on which parts to move. This isn't "do it all at once" kind of situation. .Net 1.1 is still there, works, and having .Net 2.0 doesn't break that.
.Net 2.0 are in things that use ASP, ADO, and so on. So, if you have core frameworks that are based on the core libraries, they probably can be very quickly ported to the new language. The optimum case is a recompile, retest and it's done. So, you can pick the stuff you think will mostly easily move to .Net 2.0, and port that.
.Net 2.0 road. For a really hard-core approach, pick both most critical and hardest to do first.
.Net 2.0. For example, you really may want to use those spiffy Web Part APIs on your portal, but get your old stuff running first.
There's a couple of strategies here. Firstly, you can go for low-hanging fruit. Many of the changes in
The other approach is to take a risk-aggresive approach. Take the most critical pieces you have (say, like a framework or a library that a lot of apps and so on use) and port those. Concentrate on that effort, because until those ports are done, you most likely can't move forward anyway. Repeat this approach until are well on the
Given your large code base, I think it is best to get the old code working first, as tempting as it may be to use all the latest features of
But, the key here is to pick those libraries, applications, etc. and work in stages. I don't think a big-bang migration will work very well here.