Back to the channels

#java

1. 00:15 Square I'm replacing a JSF frontend with React for a semi big java application . What approach do you feel is more natural and easy to work with: A) npm/webpack(possibly Vite later) or B) frontend-maven-plugin?
2. 00:24 Square Claude answered it.
3. 00:33 * arkadio joined #java
4. 01:04 Square Anyone using "Jetbrains Gateway" for local container development? I wonder if it creates a directory called '~' in your home folder? Feels like a bug to me.
5. 01:48 * hwpplayer1 joined #java
6. 02:10 * kcomhnall joined #java
7. 03:59 * ferdna joined #java
8. 04:23 sirecmg man jsf is the worst ><
9. 04:23 sirecmg it is probably easy to work with jsf with claude though
10. 04:23 * lain6141_ joined #java
11. 04:30 * Zapek joined #java
12. 04:37 * deavmi joined #java
13. 05:03 deebo don't do react, do htmx or something :)
14. 05:03 deebo unless you're 100% client side, having to deal with SSR is pure satan
15. 05:05 * sbalmos_ joined #java
16. 05:05 * jjj333_p joined #java
17. 05:16 Square deebo: SSR?
18. 05:17 Square oh server side rendering?
19. 05:17 deebo server side rendering
20. 05:17 Para Server Side Rendering. Acronym made to justify doing stupid things; not really the concept as a whole but more of a "I want to run React on server because durrrr".
21. 05:18 Square deebo: I thought react didn't ship with SSR out of the box?
22. 05:18 Square I've seen the horrors in Next.js
23. 05:19 deebo depends on how you serve it
24. 05:20 Square For vanilla React it's not a problem right? This is just a low traffic admin application.
25. 05:20 deebo nextjs does it ok, i'd actually use it isntead of anything self bootstrapped from a template that will be obsolete next week
26. 05:21 deebo we have a 10+ year big react+ssr project and the constant upgrading is the biggest hassle, on nextjs there would at least be more help and possibly tooling
27. 05:24 deebo unless you're working on it at least weekly for the foreseeable future, don't roll your own webpack/whatever/etc build, just use nextjs
28. 05:24 deebo especially for a backoffice app
29. 05:25 deebo antd + nextjs is a good way to go there (we run a few of these), not sure if htmx has anything comparable for components
30. 05:27 comrad Para: SSR is the proof to me, that we just went to hell and back. First everyone wanted to render on the client, now it's all back to server rendering again.
31. 05:28 Para And the reason of course was that wait, this is actually slow when we take into account all the necessary things!?
32. 05:29 Para With a good helping of forcing one trick as solution for all of it.
33. 05:29 Square deebo: I don't see the need for server side rendering for a low traffic admin application.
34. 05:29 Para All software could be made with just a single megafunction.
35. 05:29 Square now you sound like a functional programmer =D
36. 05:30 Square deebo: (but maybe I'm wrong)
37. 05:30 Square SSR is for a snappy feel was my understanding.
38. 05:31 Para At least part of it is/was motivated by Google's millisecond golfing.
39. 05:32 Square wdym? I think I've missed this. For ad brokering?
40. 05:32 Para It used to be that Google's indexer refused to render more than one full redraw, and people made shit frontends which would redraw wildly during initial load. Google also timeouted the process quickly (remember when CPU was expensive?), and site rank algorithm would use snappiness as driving factor.
41. 05:32 Para There was endless papers and blog posts about how small ms loads would lead to tens of percents of higher sales.
42. 05:33 Square ah ok
43. 05:34 * Cae2 joined #java
44. 05:35 Square So none of you use "Jetbrains Gateway"?
45. 05:35 Square ...what do you use instead?
46. 05:36 Para ...have to look up first what that even is.
47. 05:36 Square It's Jetbrains Dev Container take is my understanding
48. 05:37 Para Oh I don't use Dev Containers, those are shit.
49. 05:37 Square haha
50. 05:37 Para The idea is OK but so far I've only seen failures and even worse build setups with them in projects than what they are supposed to fix.
51. 05:37 Square So you let the agents have free reign on you system?
52. 05:37 Square your*
53. 05:38 Para I generally don't let agents run things :) I do per-command verification, and most often tool calling agents tend to deteriorate on themselves as well.
54. 05:38 Para The issue is that the "thinking" process does 20 mistakes to get 1 result, and while this could be ignored, it also means 20x cost. So don't let them think.
55. 05:39 Para Also I think at least Anthropic's models are deteriorating in general. I'm fairly certain e.g. Sonnet used to be better half a year ago than today.
56. 05:41 Square Ah interesting perspective. I've been so non curious about agentic coding so I started using it this week for the first time.
57. 05:41 deebo for anything actual user facing with SEO needs, you need SSR (still), for backoffice tooling we just have "use client;" in every component so it only works on the client
58. 05:41 Square I thought the hole industry was diving into it
59. 05:42 deebo i just give claude/copilot access to IDEA MCP and if it want's do something stupid like download jars and decompile them i tell it to just stop, and sometimes it even listens!
60. 05:42 comrad that's what their marketing wants you to think ;)
61. 05:43 deebo it's a good tool when you need to do something in bulk, or repeat the same thing over many repositories, but anything "new" still needs a few passes to clean up
62. 05:44 comrad i am lousy with webdesign, but claude did some nice sketches, so i iterated over them. not agent, but generated and tried.
63. 05:44 comrad i had so add some more prompts to stop it from remaking the whole thing, just to move a button more to the left
64. 05:44 deebo we did a few spring boot 4 upgrades manually with verification of proper configurations etc, then just pointed claude to the repos with the PRs: "look at these PRs and do the same upgrade in this repo"
65. 05:46 Para Square: On paper agentic workflows are pretty neat, it's just that the tooling seems to be focused on making an expensive mess.
66. 05:46 Square haha
67. 05:46 Para And yeah, using LLMs to generate some corners/bits which you then massage manually into shape is a pretty good way to utilize them.
68. 05:47 Square I must say this port of JSF -> React is going like proceding like a steam engine train.
69. 05:47 Para Are they even concept compatible?
70. 05:47 Para I'd probably create an extensive suite of E2E tests first.
71. 05:48 deebo yeah for agentic stuff you need tests to make sure everything actually works
72. 05:48 deebo and not tests by the agent itself :) it can create some insanely stupid tests
73. 05:48 Square I feel we've mad a big win just by cutting off the jsf dependencies. It's holding the app back in many ways.
74. 05:49 Square it's quite a simple app, but you know how things were back in the days. Tons of code for even minor things.
75. 05:50 comrad Square: that Port is a whole Concept Shift, you pretty much can start over as well
76. 05:50 Square I haven't inspected the end result yet, just checkpoints along the way. You guys don't seem convinced so maybe I'll get a sad surprise in the end.
77. 05:51 comrad but jsf is not so bad, if you know why you want it
78. 05:51 Square comrad: "that Port"?
79. 05:51 comrad Square: JSF -> React
80. 05:52 Square ah ok
81. 05:52 Square Sure, JSF is page based, React is SPA by default.
82. 05:52 Para Remains to be seen. It might work.
83. 05:52 Para The real cost is always in maintenance.
84. 05:53 Square I hoped I be able to massage away the quirks that comes with it.
85. 05:53 Square I imagine you can do React more page based too no?
86. 06:00 deebo you can
87. 06:00 deebo but no one is going to get a malicious jsf.jar of a 5 year old version in mavencentral
88. 06:01 deebo someone is going to get a malicious dependency of dependency of dependency in npm, and you're fucked if you don't have alerts etc set up for these :)
89. 06:02 Para Hm, that perks an idea. I wonder if Maven plugin ecosystem would have some kind of killswitch for downloading too old deps.
90. 06:06 deebo does age directly matter?
91. 06:07 Para There probably could be a good curve of CVE rating + time.
92. 06:23 Para OWASP has a config flag for plain CVE score.
93. 06:26 * iusr joined #java
94. 06:33 * node1 joined #java
95. 06:36 * MonsterAbyss joined #java
96. 06:42 * iusr joined #java
97. 06:56 * Krusher joined #java
98. 07:20 * MonsterAbyss joined #java
99. 07:22 * iusr joined #java
100. 07:23 * phlox joined #java
101. 07:25 * phlox joined #java
102. 07:35 * gas51627 joined #java
103. 07:40 * MonsterAbyss joined #java
104. 07:56 * Adrian35 joined #java
105. 08:11 * MonsterAbyss joined #java
106. 08:12 * lain6141_ joined #java
107. 09:15 * ta71 joined #java
108. 09:25 * iwtga joined #java
109. 09:51 * Perflosopher0387 joined #java
110. 09:52 * OmniRadix69891 joined #java
111. 10:03 * marcel1 joined #java
112. 10:10 * MikeBux joined #java
113. 10:24 * OmniRadix69891 joined #java
114. 10:43 * node1 joined #java
115. 11:11 * jamezp joined #java
116. 11:47 dreamreal morning
117. 11:49 julemand101 hi
118. 12:10 * Adrian35 joined #java
119. 12:30 * xv8 joined #java
120. 12:50 * pebble joined #java
121. 13:03 * Betal joined #java
122. 13:21 Bombe Good morning.
123. 13:33 dreamreal l'shana tova
124. 13:43 * ramontjunior joined #java
125. 13:44 sbalmos yarr
126. 14:17 * node1 joined #java
127. 14:46 * henbruas joined #java
128. 14:51 * MonsterAbyss joined #java
129. 15:12 * akaWolf joined #java
130. 15:17 * akaWolf joined #java
131. 15:21 * LtHummus joined #java
132. 15:27 * svm_invictvs joined #java
133. 15:37 * m2 joined #java
134. 15:58 * Aedil joined #java
135. 16:14 * dinomug joined #java
136. 16:15 * node1 joined #java
137. 16:40 Para I wish you all 3 parts grapefruid flavored soda, 2 parts white grapefruit juice, 2 parts gin, half part fresh lemon juice, splash/dash club soda, slice of grapefruit for garnish or whatever else might be your thing this Friday.
138. 16:58 * magla joined #java
139. 17:47 Bombe That sounds nice.
140. 17:49 Para It's the recipe for Finnish long drink, "lonkero". Drink mix which was invented for the Summer Olympics 1952 and is to this day really super popular over here.
141. 17:52 * hwpplayer1 joined #java
142. 18:09 * kcomhnall joined #java
143. 18:10 * m2 joined #java
144. 18:11 * deavmi joined #java
145. 18:15 * deavmi joined #java
146. 18:18 * mindCrime joined #java
147. 18:31 Bombe Oh, that’s what that is? I always assumed it was either a beer or contained copious amounts of Salmiakki.
148. 18:34 Para We like to scare tourists with Salmiakki.
149. 18:34 Para Once we've slapped them good enough, we'll give them the pleasant stuff. Or Gambina.
150. 20:22 * mindCrime joined #java
151. 20:30 * simon816 joined #java
152. 21:47 * Adrian35 joined #java
153. 23:35 * kathadris joined #java
154. 23:49 * mindCrime joined #java
155. 23:51 * hwpplayer1 joined #java