The wheel of reincarnation is a common phenomena in computing. So is forgetting what was learned on previous trips around the wheel.https://twitter.com/tazsingh/status/977864328199188480 …
-
-
right, the wheel turns and the past contains both positive and negative lessons. Often things that failed in the past can serve as inspirations for today. Anybody who is unaware of past turns or assume they are irrelevent is at a disadvantage.
-
Similarly, anyone who rejects divergences from past turns on the wheel is at a disavantage. Part of the way that these divergences happen is the ignorance of youth. Rather than bemoan it, we should ask why their formulations are working better than the last time.
-
The answer to "why" is almost always a context shift that can be used to revisit assumptions and build better on top of what we already know. The brashness of youth teaches us a lot and does far more good than harm.
End of conversation
New conversation -
-
-
In a more perfect world. Obviously Allen is responding to the real experiences where not enough remembering ...wastes prior learning.. Still, awful lot (and ever more) to sift through, to find what to remember, and what to selectively forget...
-
Experimentation becomes engineering becomes blackboxes. We aren’t very good at documenting worked out, well engineered solutions so they can found and reapplied in the future.
-
The RFC process used by Rust (and Ember, and React, and Yarn) does a great job of capturing understanding at the moment a decision was made, and also at mandating a certain amount of archaeology.
-
Since these communities are the targets of the "forgetting lessons" critique, I want to gently push back. Instead, I think we should cheer on these kinds of processes and help those interested in learning from the past do so.
-
I certainly wasn’t intending to critique any specific communities. Just a thought triggered by a tweet. I absolutely will cheer on such processes and encourage them to more broadly disseminate their experiences.
-
Concretely, are you thinking of specific prior art for retaining high fidelity source information across many-stage compilation pipelines where the steps are built by different people, and where the output is a single file (like an executable binary)?
-
Well, a deep literature search is always a good starting point. Also, Eclipse and VisualStudio are both examples of systems with an extensible language services API and plugin architect.
-
I work on these kinds of systems and do literature searches, but: 1. Papers are very often not available without significant $ 2. Papers are very rarely (but not never) relevant to these software engineering questions 3. Visual Studio is closed source
- 5 more replies
New conversation -
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.