Metajournals are trying to do it, with an "Article" overlay to data/software publication : http://metajnl.com/. IPOL also tries to publish image processing software: http://www.ipol.im/. And there is also some work to interlink science assets (more than articles) on http://linkedscience.org/. \ I welcome publication models not based on "PDF standard" (ie an electronic clone of the centuries-old ink&paper text article), but simply sharing data/software is not sufficient. I think it is important to have an editorial line and a peer review, otherwise anything online would qualify as scientific communication.
Moreover, software and data is hardly sufficient by itself; it needs to be explained, documented, illustrated and discussed to be more than a "harddisk dump".
Which is good as a way of 'settling information' into a medium but an awful way for today. It's the equivalent of stone tablets compared to what's available today
It would be much better if the product of research could be shared in a more flexible format. For example, don't get me started on reading two columns PDFs in a computer screen.
Flexible formatting, attachable metadata, reusable "source code" (that could be an excel spreadsheet, a txt list of data, or source code per se), redrawable graphs, etc
We need a faster, more flexible, albeit official way of sharing data. Yes, there are ownership issues, IP issues, etc. But maybe having a more easily searchable/traceable system may help with this.