Stop Adding New Features to RCs!

Helllloooo, Linden Lab!

I've been a software project manager. I know how much you programmers want to put new features in the software. I sympathize, and I WANT the new features.

But please, please...STOP PUTTING NEW FEATURES IN RELEASE CANDIDATE UPDATES!

An RC is supposed to be ready for release. That's how it gets its name. Because you can never be sure you've caught all the bugs, an RC is given a test period in the "real world". That is all just as it should be.

THE ONLY THING YOU SHOULD DO WITH AN RC IS FIX BUGS. Adding new functions carries the risk of introducing NEW bugs...and that way lies a never-ending cycle that may lead to the RC *never* getting released.

1.20(7) is a prime example. New Search. New function in chat and IM to open avatar profiles from their names. Skinning.

And what happened? Numerous crashes on login, failed hovertext prices, failed transactions. Very, very basic stuff that should NEVER happen in a software version that's considered "almost good enough to make it the main viewer". The problems were so bad that the version was pulled and we reverted to RC (6).

Let's have a little more discipline in our version control, please?

as much as i like some of the new feautes, i have to say, nicely said.

I guess that things are going pretty well in balance, if the biggest complaint is that SL didn't come up with new names for prereleases after putting in the het grid.

Just because LL uses the jargon of the industry doesn't mean that terms have the meaning one would normally expect. (cf. [RESOLVED])

Hmm jys call them all first look or beta viewers and problem resolved

An RC should require an entire feature lock, including accepting that it will get released with certain bugs unfixed. If a fix is available in an internal branch, it has to be carefully considered whether adding it is worth it, or if it can wait until the next update.

Any change has the potential to ruin the entire QA process (they supposidly do have one) by introducing untested code that could affect something entirely unrelated.

Three simultaneous viewers might work best:
* main (last released viewer)
* release candidate (locked branch, only showstoppers get merged in)
* pre-release (active branch, inherently unstable, currently the RC)

Pre-release would always be unstable since it constantly has new - largely untested - code merged in, but at some point they can decide that "bug fix X" or "new feature Y" has proven itself, doesn't introduce new problems and mark that as "RC ready". When they have enough stable changes, they can branch off an actual RC with only those changes.

RC testing wouldn't involve testing code changes then anymore (that happened in the pre-release), but simply testing the code changes that resulted from the merge.

(And dogfood their own staff.... Lindens currently download the RC at the same time we do while they should be using it in advance of us for additional testing. If they can't do their work cause they crash constantly or some necessary feature breaks again, they'd know it needs more work before subjecting residents to it)

I wonder how much the Linden staff uses SL to do their work other than the part that absolutely requires the use of SL.

Saying you can't put new features into the next version kind of negates the reason for having a next version. Sure, we all want bugs fixed. But there are many, many groups in SL working on projects simultaneously. It's not just 2 guys who either fix bugs OR make new features. Bugs get fixed, new features get added, and yes, sometimes new bugs get introduced. It's just how things work, no matter who or where. Do you know how many bugs Service Pack 3 introduced to Windows? If Microsoft can't get it right, why do you expect LL can?

That's not really what Lindal suggested though.

The idea of a release candidate is "this should be ready to be the next main viewer but we want to offer it as a preview to make sure there's nothing we missed" so they release a RC0, people test it and find a bug that makes 1/3rd of the people who test it crash.

RC1 fixes that bug and that bug only so unless the fix is flawed you *know* for a fact that you have a more stable release because you didn't add anything else and testing of RC0 proved the stability of anything but that one bug.

You can add all the new features and bug fixes you want from version to version, but once you label a branch "release candidate" your only aim should be stability as-is. Not general bug fixes, not new features, those can wait for the next version.

Hon, that's the generally accepted use of the term, not the LL use. For Linden Labs, Release Candidate means they are adding and testing new features, AND fixing bugs, all at once. And yes, sometimes adding bugs.

Linden Labs has chosen to use Release Candidates in this manner because..well..they can. It's a bit easier, I'd imagine, to develop a main candidate and a release candidate, than it is to be working with a main candidate, release candidate, and a test candidate.

First Look is used for testing the addition of huge feature add ons, before they're really sorted out and all that. Windlight for instance, was in First Look a few times and had more problems there, and was not as feature-full I don't believe.

Using release candidates as LL does, in my opinion, is fine, as long as the main candidate is generally stable.

Only if you care about making the actual release as bug-free as you can get it, which I don't see much evidence of.

.

The concept of an "RC" client has changed quite a lot recently, I would say.

Basically the RC is the main client now. You are foolish if you choose not to use it. Main client releases are extremely infrequent, and in some cases (e.g. the initial Windlight release) there are months and months during which you will not be seeing the latest developments and everybody else will.

The fact that RC clients are claimed to be "experimental" does not mean anything more than the usual "Web 2.0" thing of keeping something in perpetual beta. Everyone saw through that particular bug-excuse years ago.

I would agree there. I mainly use the RC client. Matter of fact 1.19 has some major problems with textures for me that make it nearly unusable. I use 1.19 only if there happens to be a bad realease that crashes to much then I use the normal client as a fall back only to find I miss the new features. haha.

*bump*

The "RC viewer" will get renamed to the "beta viewer" according to the blog.

And just for fun the "beta grid" will get renamed to the "preview grid" because it would be less confusing.

*thinks*

So we'll use the main and beta viewer to connect to the main grid, and a preview viewer can be released specifically for the preview grid but the "preview viewer" isn't the same as a First Look viewer because that one connects to the main grid.

Yes, that's much simpler than simply renaming the RC to be the "preview viewer" .

Never mind the adding new features ! Stop taking away the ones we already have!
When did the ability to "touch" objects while in mouselook go away ? I'm usualy 3rd person so I didn't notice it at first.

It hasn't? I'm in RC8 as I'm typing and can quite happily touch objects and get the expected response in mouse look....

Think of all the old explanations that will be wrong.

Old notecards. Old forum posts.

Ick.

I suppose we will be left with inaccurate forum titles.

Double bump

Its so crass that in todays technology and sposedly from a modern company their quality testing and releases are being done in world by US !!

Surely some form of pre testing is done with loadings similar to what the main grid would have.

Designers and builders are quite rightly up in arms about this.......wouldnt you be if you just lost work by suffering constant crashing......all because some numpty has not bothered to pre test a new script will not clash with another.

Residents are up in arms coz of the extrra lag being created by the new features, dont try saying these dont cause crashes............Its truly about time LL got thier act together and put SL into the place it should and can be..............a world leader.

it seems LL dont learn from these mistakes either.........stable releases being replaced with unstable ones is a receipe for disaster........


Grrrrrrrrrrrr now im on my high horse..........i cant get off it.......this is my major beef with LL.........when they have professional programmers telling them theyre being stupid.....and they still dont listen, it makes me wonder if these comments ever gets filtered back to somone who does give a fook.it wouldnt seem so huh

Breath in.............breath out.............calm,,,,,,,,serenity........peace..........breath in.hold......then shout LISTEN TO US LINDENS

Thats strange I haven't been able to open my doors in mouselook in RC8 or RC7 !!?@?

Now there are having a go at me personaly then !!!

Leave me alone I haven't given in to the urge to shoot/kick a Linden when I see one in world, stop picking on me !

i dont know if this is the htread i should be on but i had to update the client and to be honest it really sucks big style i see peopelas balls of mist and the lag is aweful also the shininess on builds doesnt work i can have my graphics set as high as i used to two updats ago and i just logged in to find myself as a ball of mist and my builds too ten minutes to rezz this is getting to be very painful and i wish i could go back to windlight i found nothing wrong with it at all i cant even look a t the awesoem linden water anymore this sucks

Some scripted attachments that take controls can remove that ability. I did it accidentally myself in my line of collars, had to fix the issue with an update (1.10 if you haven't updated, do so )

Heh heh, Google reports 2.880 occurences of the phrase "beta grid" in the secondlife.com domain.

All of which will be misleading now.

Thx I found it It is a combat HUD which covers the aiming point and some weird camera stuff too. I'll send the makers a note to let them know.

But then how will I be able to tell the professional software development houses that actually understand the dev process from the hacks who don't? I need some way to distinguish which software products I trust enough to incorporate into my business model (hint: SL isn't one of them).

IMHO, they need to stop putting in new features, period. At least, until they have all the other problems fixed. They should have a stable platform, both client- and server-side before they start adding new features.

As of now, the client is unstable (lots of crashes still) AND it's clear that their server platform sucks, as there are all sorts of problems.. like the ones that are happening RIGHT NOW:

"We have reports of sporatic unavailablity of some services such as
login not working, transactions failing, and land store down.
We are aware of the problem and are working to bring these services
back online as quickly as possible."

Oh, and you didn't mention the broken searches and all the other problems that occur anytime the resident numbers get up there.