Precedent suggests it's OK to make thumbnails (highly transformative, and not competition), and that it's also OK to hotlink.
From wikipedia:
> conduct was excused because the value to the public of the otherwise unavailable, useful function outweighed the impact on Perfect 10 of Google's possibly superseding use.
> Moreover, in Perfect 10, the court laid down a far-reaching precedent in favor of linking and framing, which the court gave a complete pass under copyright. It concluded that "in-line linking and framing may cause some computer users to believe they are viewing a single Google webpage, [but] the Copyright Act . . . does not protect a copyright holder against acts that cause consumer confusion."
There might be some cases where ajax calls cross the line into infringement.
Caching is easier. Once again, Google smacked down a plaintiff:
http://en.wikipedia.org/wiki/Field_v._Google
> Courts usually do not require a copyright holder to affirmatively take steps to prevent infringement. In this case, however, the court found that the plaintiff had granted Google an implied, nonexclusive license to display the work because of Field’s failure in using meta tags to prevent his site from being cached by Google. This could reasonably be interpreted as a grant of a license for that use and so the courts held that a license for Google to cache the site was implied because Field failed to take the necessary steps when setting up his website.
If the server allows caching (see meta tags), it seems OK to cache it. But it depends (Google had to rely on meta, because they were caching everything ... if you specifically cache a website, you might want to see if hey have a TOS).
I'd just be a little careful in how you transform the data. Also, there's API rules.
Really, you'd need to see a lawyer.
Scraping these days may more often be a means of extracting data to an offsite cache/database, but ad-hoc on-demand scraping is consistent with the definition.
Regardless, this doesn't really matter. I don't see how any restrictions on "scraping" can possibly be enforced. Screen reader programs that allow visually impaired individuals to browse the web are "scraping". Mobile browsers that reformat pages on the fly to optimize their display on a small screen are "scraping".
Ad-hoc user-initiated scraping of this sort is really no different than just browsing the web directly. So whether one wants to argue that "scraping" is or isn't happening here, it hardly matters to whether this is ethical, and should not matter to whether it is legal (although on that topic, I'm guessing different judges would see it differently).
Duh, of course it doesn't. Scraping is taking data in one format from there and presenting it in a different format here.
If not, I have an interesting idea for an offline travel app :)