Publishing at a venue without ACM/IEEE brand?

Not being affiliated with an organization such as the IEEE or ACM is not, on its own, a bad sign. For example, what is currently the IEEE Conference on Computational Complexity has decided to end its affiliation with the IEEE and go solo; the Symposium on Computational Geometry likewise just left ACM. STACS is also unaffiliated.

It's not great if the proceedings are only distributed on CD. On the other hand, if you're already going to be preparing a journal version (and you should!), that will completely supersede any conference version, anyway. I would never read a conference version ("extended abstract") of a paper if a full version was available: in my area (theoretical CS), the conference version generally has the proofs missing and the peer review isn't very rigorous.

STACS publishes proceedings through Dagstuhl, which is basically a computer science conference centre in Germany and has close links to DBLP. In this case, copyright is retained by the authors. Note that, regardless of who owns the copyright, most CS conferences won't accept papers which have either appeared at another conference with published proceedings or have been submitted to such a conference. You don't get to double-dip your papers.

I think that is very field-dependent. In my field there are two conferences that are not affiliated with IEEE. They are both small, highly-specialized, and very prestigious to get your papers in. One of them does publish the proceedings in a book with a highly-reputable publisher, the other one publishes the papers online only. Still, they are well-known and well-regarded in the field.

The bottom line is that you need to know your field. It certainly might be a red-flag, but not necessarily. The conference is as strong as its PC and its acceptance criteria.