2 ms·
I agree with you on both counts. One of my first jobs was wearing a build engineer hat building software in a completely clean-room environment (no network, no
by gjvc 6y ago
I agree with you on both counts.
One of my first jobs was wearing a build engineer hat building software in a completely clean-room environment (no network, nothing), using completely vanilla installations of operating systems and compilers.
A few years later, when I moved to Java, I was shocked to see people download .jar files from the central Maven repository without a care in the world, use them in development and do the same process for production. A few years after that, npm came along and made it normal for Javascript. By this time there appeared to be no stopping the madness, and incidents like this https://www.theregister.com/2016/03/23/npm_left_pad_chaos/ https://www.theregister.com/2016/03/23/npm_left_pad_chaos/ seemingly failed to give people pause.
CI/CD as practised today contains huge risks on what is ingested to a given corporate infrastructure -- millions of lines of code from external sources of differing security levels. But instead of all that in plain sight, a single inaccurate article based on speculation can get people worried about something else. (related, but as you say, a very long way off from the most targets potentially most vulnerable and valuable-to-an-attacker). Crazy.
- burntoutfire 6y agoI'm guessing companies assume that whatever code gets written in-house will have more vulns that some widely used open source project... Not to mention the cost to re-implement the world (sans the OS and compiler) would be astronomical and put most projects into the "not worth it" category.
- gjvc 6y agoI'm not talking about not-reinventing-the-wheel, I'm talking about making network calls by default, to fetch dependencies. This makes review of such things seem obstructive. As ever, convenience and security are opposing forces.