July 10, 2016 at 05:50Let me go, Meteor
It is not easy to give up illusions, but the time has come to draw a line. Maybe someone will fit at the crossroads of choice. Well, then I will show my grandchildren. This story is already 20 years old. Summary: frameworks - never again; there is an exit.
You can immediately go to the bottom , or with a preface: why is the Meteor at all?
Started on Delphi under the flags of RAD and VCL . Two suitcases of books - bought everything in a row, like a maniac. It's nice to know that some learned to code in the Winsyntax editor .
Years passed. He got a job as a webmaster in the IT department. In addition to shaking the corporate website in PHP, I met with the SharePoint monster: widgets in C #, InfoPath forms, a gateway to HP ServiceDesk in Java. And I really wanted to raise a website about my typewriter .
Like SharePoint, but so hello in PHP, Drupal. It was love at first sight. Delphi for the web. Register of modules - treasure box. The basic functionality can be customized with the mouse. Collective mind to solve technical issues. I got sick myself and infected the author of LiveStreet :
Work on the engine (still just a universal engine, without reference to blogs) began almost two months ago, after a boost of energy from communicating with an advanced PHP developer.
Then there was Mamba. I learned about Highload and Big Data, about the costs of developing and maintaining a big-big bike. Then he tried on the role of a major expert in web technologies in a fat startup: the framework will save us from any troubles, in one go I’ll raise the super-duper project to the Maldives. Django Python is charming, but no. Maybe Yii? Well, what for PHP in general. V8, NPM, CoffeeScript, Jade, Express - expanse. The roof went from SocketStream (wow, still alive), thanks to the author for the tip - hi, Meteor. Duplicated one or two, as it seemed to me. Today, I have already decently accumulated full-time practices.
And with all this we try to take off
Suddenly realized the cynical truth. I chose frameworks to compensate for the lack of resources for a quick start, but this is a conditional value when running a long distance. A framework by definition has redundant functionality - you need to cater to all-all-all use cases. "The more we rent out, the better." No matter how beautiful the framework may seem, the story will be repeated every time. Our business goals diverge.
Recall the terrifying intricacies of the data scheme under the hood of Drupal, where dumb caching saved ( Alexander Solovyov spoke about this in his report on optimizing modnaKasta).
The framework takes responsibility by limiting the choice of third-party solutions. Initially, dubious decisions interfere. Correct decisions become obsolete. The meteor changes and packets drop: AutoForm, CollectionFS, Jade, IronRouter, crafts from guru @arunoda. The lifetimes of packages dependent on the framework are significantly inferior to the libraries for the platform. And here is a specific ambush: there are thousands of lines of branchy left code on support with obligations to update dependencies for new versions of the framework.
Atmosphere - the official register of packages that stores tons of idle junk. And it no longer collects statistics of installations, although the mechanism can be built into Meteor (I recall Drupal). From the "wealth of choice" eyes are scattered. The situation is aggravated by the rejection of free hosting of demos on * .meteor.com - does the launch of Galaxy block the possibility of further external investments? Well, OK.
And this atmosphere of yours slows down the whole way (although you can speed it up, see Fastosphere) - nobody cares.
About the redesign of the documentation . Previously, this was one page with anchors. To whom did she interfere? Links from Google no longer work (guys apparently are not aware that hypertext is the integrity of links in the first place). Under the howl of criticism, after some time an internal search appeared, but keep it to yourself (moved to devdocs.io).
The more responsibility the framework takes on, the greater the technological gap in the basic packages. Selected examples: HTTP / 2 , Mongo , Node . Build speed increased noticeably with the transition to version 1.3 (the long-awaited support for NPM), but it's still naturally trash.
The cornerstone of productivity (yes, yes, yes, I speak only for myself) is sent to landfill. JS bit the bit, alas. The language should be simple at a basic level. But the framework is bound to follow trends, yeah. Really met the demand in vacancies: aversion to CoffeeScript. See how beautiful React is in CoffeeScript.
Does GraphQL close LiveQuery scalability issue? Wait and see, I only need it today. While Horizon is ahead of Apollo . Plus, Victor Grishchenko criticized DDP at the HolyJS conference . They say there is no id version in forwarded packets. And this means that if users go offline and return after some time, then "optimistic updates" (aka latency compensation) will not work and the data will be lost.
Favro - smart , but the source code is not available. Still there RocketChat, but there pechalka :
I believe that RocketChat will outgrow Meteor. We have been developing many workarounds to Meteor's limitations. As we develop more specific code to deal with our edge cases, move to ReactJS, create our own plugin / package system and Meteor 1.3 get more compatible with NPM, I can see we getting into a point when de-meteorising Rocket.Chat may be easier and necessary.
I did not find anything else interesting .
By the way, here are alternative sets on the knee with a jigsaw without a claim to be called frameworks: Meatier , lovli.js .
Just now I learned how it was conceived :
Meteor is designed to take the technology that used to be exclusive to well-funded frontend teams at Facebook and Google, and put it in the hands of everyone.
The meteor is increasingly moving to the Facebook technology stack, apparently wants to merge in ecstasy. Amen!
Total
You need to walk to the horizon to see a new horizon.
PS: one more has seen .
PPS: why frameworks are evil .