5 ms·
This is very well written.
by nbevans 10y ago
This is very well written.
- cm3 10y agoIt might be dangerous to click on it, is there a tl;dr or text extract of it?
- dan1234 10y agoYou could probably use an online pdf-to-text converter[1] or view the PDF from within a virtual machine. [1]http://pdftotext.com http://pdftotext.com (no affiliation, first search result)
- cantagi 10y agoIt's OK - Github converts it to HTML. However, GCHQ could politely ask the NSA to force Github to give them the usernames and IP addresses that viewed the repo. Maybe use Tor?
- coroxout 10y agoIt didn't convert it for me. Oops. Well, they know who I am anyway. Hi again! (Edit: If you click on it from the repo file list at the top it converts it; don't do what I did and click on the link to download at the bottom of the readme)
- mseebach 10y agoI don't understand why everybody assumes GCHQ and NSA are completely inept amateurs. Getting a list of usernames and IP addresses from GitHub would have an insane signal to noise ratio - after all, there are plenty of perfectly legitimate reasons to want to read this document. It's much easier to go to HN and grab the usernames and IPs of people you can see taking a rebellious position in discussions about these organisations. Also, why would you assume that showing any interest whatsoever in Tor isn't going to land you on many more watchlists than reading a public PDF in GitHub?
- chinathrow 10y ago> Getting a list of usernames and IP addresses from GitHub would have an insane signal to noise ratio Or you could assume that they have that information already. Which I deem highly likely - private repos are a must-target for any intelligence service out there.
- nxzero 10y ago>> ..."GCHQ and NSA are completely inept amateurs." Assuming anyone is inept is inept, since even the inept get lucky.
- Spearchucker 10y agoActually I wasn't that impressed. I found it hard to focus on the text - there's a lot of government speak (like "ways of working") in there. Also, the author has a fixation on Conway's Law ("Conway" appears 40 times in the document), and everything else is really a spin on Wardley's evolution curve. Both are useful, to be sure, but there's nothing inherently profound or new here. This is (should, maybe?) be targeted at CxOs of big corps, and I think has some value there - if only it were more pithy (which is not the English way). There is also some irony in that they're saying "do away with big planning upfront" in the same document as "be big-data-driven". Big data needs big planning up front if it's going to be meaningful.