4 ms·

> On the other hand, you risk ending up in local optima. This would only be a risk if the difference between the local optimum and the global optimum is large,

by pron 8y ago

> On the other hand, you risk ending up in local optima. This would only be a risk if the difference between the local optimum and the global optimum is large, but there is no evidence to suggest that is the case (and a lack of evidence for a big effect is usually evidence against it). To be more precise, no one has been able to find a drastic productivity boost among apparently reasonable alternatives, so either there's a plateau or no one has found the global optimum. Either way, as things stand, there is no such risk. A risk that does exist is failing to create a viable development environment or simply not having the resources to even attempt it. As others noted, this is a big risk (or an almost guaranteed failure) for small organizations, but hardly one for a very big one.

yayana 8y ago
The overall business impact on using php vs any other language is clearly huge. No other language is split into a separate market for developers and VC capital. JS hate couldn't stop its use completely for any company that needs a website, Java will likely be running something whether you develop your API in it or not, only php hate combined with php's only role in the stack being high but substitutable can eliminate it from businesses and developer mindset with prejudice at some companies and make it the dominant language at others.
humanrebar 8y ago
On the contrary, certain kinds of technical debt are only paid off in full, not in increments. Supporting entire languages, for instance: experts, toolchains, dev environments, packaging systems, core libraries, telemetry libraries, etc. The large difference between a local maximum and a goal is clear. The question is whether the cost to get there is acceptable.
Nitramp 8y ago
> This would only be a risk if the difference between the local optimum and the global optimum is large, but there is no evidence to suggest that is the case Agreed, it's been elusive for the software engineering community to objectively assess and quantify productivity effects of software tool chains, in particular in comparable stacks. Comparison of programming languages or stacks is not what I was (trying to) talk about though. I have no idea whether hack compares well to $RANDOM other toolchain. I am curious as to whether a continued incremental investment in PHP made sense, compared to choosing a whole stack and propping that up to meet your requirements. Even assuming your final PHP-based stack is identical to whatever else you'd have chosen, the paths do have differences in effort. The local optimum in that context is that it's locally optimal to just add that one feature to your existing solution vs investing in a different solution that in the long term will cost you less to build and maintain. I don't have an answer to that question though, as I'm not familiar with the effort required or comparisons made. I think it's a genuinely hard call to make. Even post hoc and with all the knowledge it'd be hard to judge, as you need to discount effort by the relative risk factors.

by pron 8y ago

yayana 8y ago

The overall business impact on using php vs any other language is clearly huge. No other language is split into a separate market for developers and VC capital. JS hate couldn't stop its use completely for any company that needs a website, Java will likely be running something whether you develop your API in it or not, only php hate combined with php's only role in the stack being high but substitutable can eliminate it from businesses and developer mindset with prejudice at some companies and make it the dominant language at others.

humanrebar 8y ago

On the contrary, certain kinds of technical debt are only paid off in full, not in increments. Supporting entire languages, for instance: experts, toolchains, dev environments, packaging systems, core libraries, telemetry libraries, etc. The large difference between a local maximum and a goal is clear. The question is whether the cost to get there is acceptable.

Nitramp 8y ago

> This would only be a risk if the difference between the local optimum and the global optimum is large, but there is no evidence to suggest that is the case Agreed, it's been elusive for the software engineering community to objectively assess and quantify productivity effects of software tool chains, in particular in comparable stacks. Comparison of programming languages or stacks is not what I was (trying to) talk about though. I have no idea whether hack compares well to $RANDOM other toolchain. I am curious as to whether a continued incremental investment in PHP made sense, compared to choosing a whole stack and propping that up to meet your requirements. Even assuming your final PHP-based stack is identical to whatever else you'd have chosen, the paths do have differences in effort. The local optimum in that context is that it's locally optimal to just add that one feature to your existing solution vs investing in a different solution that in the long term will cost you less to build and maintain. I don't have an answer to that question though, as I'm not familiar with the effort required or comparisons made. I think it's a genuinely hard call to make. Even post hoc and with all the knowledge it'd be hard to judge, as you need to discount effort by the relative risk factors.