The answer depends on how much experience of developing real time games you actually have.

If you've never coded anything before, then don't bother trying to write anything ambitious yet. I know it sounds boring, but start with pong, and work up to more ambitious projects later.

I started my Project with the Main Menu and the Credits Scroller (not very important parts, i know, but I think it was fun)
At the moment, I'm writing the part to draw the map.

There is a nice book out there:
Real-Time Strategy game programming with DX

Maybe if you told us what tools and skills you have and what RTS game in
the market you would like to create similarly, someone could give you a
better answer. Goto www.amazon.com and read the reviews of books BEFORE
you decide to buy anything.

I'm afraid if you have not done this before, you should develop everything
separately. Create a test demo for scrolling, a test demo for blitting bitmaps,
etc.. I would say you won't see a *playable* game until after 6-12 months
of coding. It's more work than you think. Don't forget to start creating
a map level editor along with your game. Add another 3-6 months to your project.

Retro

Thanks Retro.
It's a Warcraft II, AOE like game.
I have done some things before.
I've been working for the last 6 months at this game (a small team) and now we're trying to accelerate the work because we think we can do something good. Until now it was just an ideea.
The bosses seam to be pleased so we got some new team members.
We have a pretty good map editor (I started with that), some bitmaps and a small part of the graphical engine.
Now we're working at this engine trying to make it faster and complete (it's drawing a lot of bitmaps and pixels on the screen after some simple algorithms).
It just seams to me that there is a lot of drawing to do and that it's gonna take much more time than I initially thought.
I don't have the bitmaps I need yet, so I must do the software flexible and adaptable to a lot of things. Is this ok? Because we're working more than we would have if we had the pictures.
Should I have a bitmap compression algorithm yet? Or I do that at the end?
Should I do any AI before I finish the engine?
And a lot of other questions...
I have answers to all of them.
But I don't know if those are the right answers.

------------------

I'm doing something similar to that, and I remember when I first started adding
units, I created my path-finding stuff first. I was able to selected single
and multiple units, and send them anywhere on the map. The two basic units
that I added were a fighting ground unit, and a passive resource gathering
unit, which are self-explanatory.

Then after the path-finding, I added AI. For the gathering unit, this means
he could go to the base, unload his resources, and he would remember the
location where he was previously at and return there, get more resources, etc..
For the ground unit, it would mean he would attack any enemy units approaching
nearby. And chase them down if they try to escape.

So, this is the basic concept of AI. Units that can respond and think all
without player intervention. And that is just the human side (you the player).
If you also create a computer player with his own base, units, etc., this is
all AI. So what I did was create the human side first so I could interact
with it, and then developed the computer side of the game using the same code.

With bitmaps, I started out using uncompressed bitmaps that were loaded from
their own file. I then switched to compressed PCX files. And then created my
own file packer program and packed everything into a single .dat file.

All game media like graphics and sound, I tried to do last. Instead, I just
substituted basic graphics, like boxes with arrows so you know what direction
the unit is heading, and simple sounds. Programming always comes first in my
mind.

And since this was the first time I have ever tried to create a game, it was
not organized. I created code where and when I needed it, which was a bad
way to do it. So right now, I am trying to clean up the old code by creating
class managers, like a surface manager, sound manager, etc.. For example, all
DD surfaces will be handled (created,destroyed,etc.) by my surface manager, so
I won't be able to get a DD surface anywhere else. Which is a good thing.

Of all the most complicated code I wrote (so far), I think it was the merging
of path-finding and AI. Sometimes I had to print out the entire code out on
paper just to track down bugs, because you can open only so many files on the
screen. I must have rewrote this section over and over again hundreds of times.

Anyway, I am far from being finish.

Retro

I'm trying to start a RTS (2D bitmaped). I have a small team and everything I need for programming.
I don't know what should I start with and which are the next steps.
If anyone can help me...

dude, if a question like "how do I start an RTS game" enters ur mind, go write some serious pong and tetris.