Any such interactions with vendors should be disclosed explicitly within the article. I wrote reviews of a couple of books for TechRepublic -- one of which is "A Practical Guide to Linux Commands, Editors, and Shell Programming", and the other of which is "The Book of PF". In both cases, I reviewed the second edition. The URIs for these two articles here on TR are:

http://blogs.techrepublic.com.com/opensource/?p=2413

http://blogs.techrepublic.com.com/security/?p=5030

They're both positive reviews. I was paid for neither. I was asked to allow myself to be quoted in the frontmatter of one of them (the Linux book), based on a glowing review I gave the first edition of that book on Amazon years ago; I was given a copy of the book as a "thank you" with no expectation of the review I ended up writing for TechRepublic. I was asked to review the other book, and given a copy of the new edition, but no requests were made about the content of the review.

The key to what differentiates these book reviews from the sorts of reviews where one gets paid for a review is that I was in neither case asked to give a positive review, there was no actual money involved, and I explicitly shared the whole story in each case, right there in the article. I have never been accused of any bias in these articles, as far as I'm aware.

I have, however, been accused of bias and (entirely nonexistent) industry ties in cases where I offered a negative assessment of a given corpporation or piece of software. This includes being both accused of having a personal vendetta against Microsoft and being a Microsoft shill, on separate occasions only weeks apart within a period of writing (more and more regularly) for TechRepublic for nearly half its existence.

The problem with disallowing all contact with a company is that you then effectively end up disallowing the ability to provide news and reviews. If all pertinent interactions are disclosed in explicit detail, preferably within the text of the article, I see no problem with things like getting a free copy of whatever is under review. Even time-limited NDAs might be acceptable, if the existence of the NDA is not disallowed as material for the article.

The more we simply disallow, the more people are going to look for loopholes. If there is a requirement of 100% disclosure, however, the reader is made fully aware of any potential conflicts of interest, and can make informed decisions about the article.

(edit: clarification)