« first day (513 days earlier)      last day (176 days later) » 

12:01 AM
well
maybe begin tran/commit will help
 
waits are just numbers
@PaulWhite there was an INEXPERIENCED RIDER on the bus today
 
so is age until you're seen by a dad
 
this rider needed to get off at my stop
she pulled the cord far too early
Erik would have turned up his nose
then pulled it a bunch more times and didn't understand why it wasn't working
I had the perfect view of this, best seat in the house
she then jumps to the other side of the bus and starts pulling that cord
still no reaction
I see the internal conflict: "Do I go up to the front and ask the driver to help?"
 
why didn't you help?
 
I put her out of her misery by pulling the cord at the "correct time"
@ErikDarling I did help
 
12:04 AM
a true citizen would have alerted her to her folly
 
she was half a bus away. Also I don't want to help the driver with his BS in any way
 
Did you say Hope that Helps
 
ok I'll make a fake story for you guys
let's do a retcon
...
she then jumps to the other side of the bus and starts pulling that cord
happens to be in the seat next to me
 
you forgot to scratch the record
this show sucks
 
I wait until the correct time and pull the cord
the stop is signaled
I lean in towards her
and in her ear, I whisper
"hope that helps"
 
12:06 AM
If he says he talked to a girl, the bus stories have jumped the shark
 
she starts crying and misses her stop
@PaulWhite I talked to a girl and she started crying, totally believable
 
> tfw to intelligent
changing the code to this cut writelog in half
CREATE OR ALTER PROCEDURE dbo.VoteInserter
AS
SET NOCOUNT, XACT_ABORT ON
BEGIN
    BEGIN TRAN
      INSERT dbo.Votes_Insert ( PostId, VoteTypeId, CreationDate )
                       VALUES ( 138, 2, '2019-06-27');
    COMMIT
END
GO
sql server is absurd
 
So the cord is now disabled until minimum distance?
 
@ErikDarling do you think that Paul has seen the classic comic?
@PaulWhite oh, did I forget to mention that? the driver started manually disabling the cord until a very short distance before the stop
as far as I can tell, this is only done for my stop
 
@JoeObbish probably not. but someone should show him.
 
12:09 AM
it took me a short while to figure that out. seemed like the bus was broken
 
OH MY GOD
THE GIRL WAS JOE
WHAT A TWIST
also brb
 
@ErikDarling wut
you used too much pre workout or something
 
@JoeObbish that thought did not appear in the transcript afaik
 
12:14 AM
Jun 12 at 23:00, by Joe Obbish
the driver started disabling the cord until right before the stop
time to watch a PG movie
what am I, Erik?
 
ah, missed that then, thanks
@ErikDarling For when you're back, would it be interesting to compare those results with HK?
 
@PaulWhite yes, but i don't have it set up. if you have a quick script to do so i'd test it.
that does bring me down a rabbit hole of natively compiled etc
i assume the begin tran/commit code is faster because of the lack of log flushes or something
 
I was thinking interop as an easier option. Also possibly more realistic
@ErikDarling no idea. Curious
 
i have no idea what that means
 
@ErikDarling normal SQL against a HK table
Same script, HK target
 
12:27 AM
ah okay
 
A natively compiled comparison would likely be mundane
Also I'm not writing that on my phone
 
why even bother having an iphone
 
Bus news when out and about
 
perfmon counters to the rescue
top result is begin tran/commit, bottom result is without
 
12:55 AM
well this is a blog post
 
I look forward to reading it
so the explicit begin tran...commit inside the proc makes a difference?
even when the proc is executed just once per batch?
 
Yeah
It's weird
 
1:15 AM
have you looked at fn_dblog?
I'm curious if the separate commands mess with entries into the log buffer
 
Not yet
Done for the night
 
 
2 hours later…
3:44 AM
Is there any explanation, that can be shared, for what OPTIMIZE_FOR_SEQUENTIAL_KEY is doing under the covers?
 
@MichaelGreen if someone looks hard enough, sure
likely enough effort involved and interest from others to warrant a question on main
 
4:29 AM
Touché. I'll type up one later. The "show what you've tried so far" section's gonna be .. um .. sparse.
 
 
2 hours later…
6:04 AM
@MichaelGreen If you can wait, Pam has promised to blog about it twitter.com/SQLGoddess/status/1144325911434555392
 
 
4 hours later…
10:01 AM
@Forrest it's odd. the begin/commit code generates more log entries and when i sum up log record length and log record reserve, they're both higher
Log Record Length	Log Reserve
66957316.0	    33251278.0
Log Record Length	Log Reserve
66915212.0	    33226340.0
 
10:25 AM
@ErikDarling Just curious, when you did the delayed durability test, did you set it to forced or add a WITH to the commit transaction statement?
 
forced
i can try the other way if you think it'd offer any advantage
 
no I just wanted to check that specific thing
 
now i'm questioning everything i did
THANKS PAUL
 
11:00 AM
@ErikDarling are you saying he shouldn't have asked that on twitter?
Honestly I think that's going to be most people's first question after reading the limited docs on this.
I can understand them not wanting to get too deep into it, but it's tough to assess something without at least a clue about what it does under the hood.
 
no
i like that twitter gives you fast access to people who might be difficult to reach
 
There must be a trade-off, else why would it not be automatically enabled on e.g. identity or any index with stats identified as the ascending key type.
@ErikDarling Oh ok. I didn't get that from "twitter is for documentation."
 
i just can't imagine that what this does in the background is easily addressed in limited character space and probably without some pictures
but hey who knows
 
A skilled technical blogger could do it I'm sure.
 
aw, thanks
 
11:04 AM
You should offer your services.
 
i have; they were declined
 
Meanies
 
yes the wound is still fresh
 
I suppose an obvious idea is to replace acquiring and releasing an exclusive latch per thread with something that allows multiple insert-only tasks to share a single exclusive latch within a suitably small time window. For pages identified as being hot spots of insert activity.
No idea if that's even close to correct, but it's what I might do.
 
they probably just added a NOLATCH hint
 
11:21 AM
> hobt_insert_flow_control_access
Occurs when a task tries to access the flow-controlled section during insert into a B-tree.
> hobt_insert_flow_control_grant
Occurs when a task gets access to the flow-controlled section during insert into a B-tree.
> hobt_insert_flow_control_release
Occurs when a task releases access to the flow-controlled section during insert into a B-tree.
 
do you think it works on a heap with an identity?
i mean it's called hobt but the description doesn't mention the ho
 
the description only mentions b-tree.
 
that's what i just said
read the transcript please
 
also the docs only talk about indexes
 
docs == ops
 
11:29 AM
@ErikDarling no you said something about ho's
NYC vernacular I assumed
 
"heaps or"
get your mind out of the GUTTER
i suppose it should be "heap as" for you
 
just the h would have done
maybe they're trying to deprecate heaps
 
i go above and beyond in all things
 
have you tried a google search for "optimize_for_sequential_key" recently
 
heaps will only be available in hekaton
as recently as yesterday evening
 
11:32 AM
I get four results. Two are to this transcript.
 
@_@
ABORT
 
I get the same
 
BURN IT ALL DOWN
 
scary huh
ROLLBACK TRANSACTION;
 
11:33 AM
i was never here
 
If someone creates a question I imagine that will be fairly close to the top of the list
As the list is 4 items
 
bing returns zero results
good job, Microsoft
 
best search engine
 
I hear it powers Windows 10 search
 
Is that why it never finds what I want
 
11:35 AM
that's cortana's south american cousin santana, which is why there's a bit of language gap
 
Or if you type too fast it just bing searches it rather than opening the program
I should not be able to type faster than the computer can understand it
 
Set-ItemProperty -Path "HKCU:\Software\Microsoft\Windows\CurrentVersion\Search" -Name "BingSearchEnabled" -Type DWord -Value 0
 
My favorite is when I search something, the thing I want is highlighted, and search results are updated milliseconds before I press enter.
 
Oh crap really?
 
@ErikDarling Did you just post PowersHell?
 
11:37 AM
it beats posting regedit instructions
 
@JoshDarnell It has many fine features, developed with careful usability testing
@ErikDarling Does it. DOES IT.
 
first, press the windows key and r, then type in regedit.exe
yes
 
@ErikDarling Is there a powershell command to do that?
 
PS fan confirmed
 
9 fraulein
 
11:39 AM
@PaulWhite Practically goes without saying.
 
@ErikDarling The evidence is right there in the transcript for the whole world (except Bing users) to see
 
the enemy of my enemy is my friend
 
ship any old shit - Satya
has anyone tried dragging a column from Object Explorer to a query window in SSMS 18
 
yeh broken
there's a q&a about it
 
> careful testing
 
11:41 AM
Horrifically broken
 
"we don't test things no one uses"
 
I heard it crashes SSMS
 
"it"
 
well yeah
ok, among many things
perhaps it's part of an effort to migrate everyone to Azure Information Data Studio
 
It's a bug fix. You can't convert a table to text. the data-type conversion rules won't allow for it.
 
11:43 AM
or whatever it is called
 
@PaulWhite :O
 
Azure Internet Data Studio?
 
There's a whole tag about at this point.
 
Does that mean if I use it I'm an Azure Internet Data Artist
 
AIDA?
 
11:45 AM
Ah. I see now.
Missed that one.
 
It was quite subtle
 
SSMS4lyf
I wonder how many people complained about db diagrams for them to put it back in
 
one person from sap
2
 
It's hard to see what removing diagrams actually achieved anyway
Just annoying for people who used it
 
Removed 500kb from the installer size?
 
11:47 AM
heh yeah
upgrading 2019 in place was a failure
 
The chief complaint I remember seeing is that people accidentally click it, and have to dismiss the little dialog.
 
@JoshDarnell why not just... remove the dialog?
 
Yeah that is slightly irritating I suppose
 
Huge value add.
I've done that once in the last year I think.
 
I hear options page space is at a premium
Not easy to add a checkbox
Requires months of careful usability testing
 
11:50 AM
@ErikDarling You're hired.
 
Does anyone remember last time SO/SE changed something and it was well received?
I'm sure there was something, I just can't recall what it was
 
new top bar
 
you jest
 
11:53 AM
@PaulWhite Seems like you could replace SO/SE with "any company"
 
@JoshDarnell Verbose error messages for string or binary data would be truncated went down well
 
But still no SSMS Dark theme. Tsk tsk
 
@PaulWhite Oh yeah.
 
One day that will come out, and there will be nothing but complaints, and it will be glorious
 
Maybe the revision history entry for HNQ entry?
 
11:56 AM
Yeah, I don't remember folks complaining about that.
Other than the lack of an entry for HNQ departure.
 
Oh! Triple backtick markdown!
That was it.
 
Haha I feel like you are scrolling through the same meta Q&A as me, only slightly faster.
 
@George.Palacios i think there was a workaround or something for that but it didn't work on the results pane so
 
717
Q: Recent feature changes to Stack Exchange

devinbThis is an unofficial list/changelog of new features and various changes to Stack Overflow and the Stack Exchange network. It is maintained by the community, while a Stack Exchange employee changes the accepted answer to ensure that the latest changes remain on top (given default user settings)....

 
@JoshDarnell yes
 
11:59 AM
2010-05-08: User profile character limit is now 3,000 characters (up from 2,000 characters).
Well thank god
 
Hitting the important stuff first
Like the tag synonyms page meta.stackexchange.com/q/329687/194636
 
SEDE must have been well recieved
 
that's a reach
 
June 2010
The exact moment SO peaked
I guess I shouldn't have read meta.stackoverflow.com/q/386324/440595 today
 
12:11 PM
ouch
 
At this point, I'm just fed up. I once believed that I was helping the site out, only to be referred to as some kind of "aristocrat" who would flog those in need of help with scorpions.
That's a great bit of imagery
Hey random question: Are the rules around comments different in meta than in the main boards?
 
@George.Palacios Yes. Meta is for discussion.
 
Maybe "rules" is the wrong word. Attitude I suppose
The 100+ comments on that post struck me
 
Even I don't delete comments on meta very often.
 
12:24 PM
The Q & A format doesn't work brilliantly for meta, but it's what we have
 
12:41 PM
well
i eagerly await jrdba.stackexchange.com
 
closed as a duplicate of op.stackexchange.com
personally I favour tagging questions beginner/intermediate/expert
 
so we'll have op.stackexchange.com and postop.stackexchange.com?
 
heh
 
the tagging thing is interesting
what goes along with that?
is it just a means for people who are burnt out to filter out beginner stuff?
 
pretty much yeah
I'm also in favour of expanding high rep/gold badge moderation powers
 
12:46 PM
quite the reformer we have here
 
well SO is quite broken
 
breaking: high rep mod advocates moar power for high rep mods
;p
 
I just think the caretakers/curators need some love
 
i'll give you my lOve Paul
i promise it's nOt a traP
 
we need to review chat access at the next room owners' meeting
 
12:50 PM
blame Sean
i only made it in here b/c he @'d me at some point
 
he's top of the review list tbh
 
i'm tired of him smearing rcsi
 
Oracle made it work somehow
 
well it's clearly what's holding back azure sqldb
 
I hear of users writing their own locking implementations just to get stuff done
 
12:56 PM
it's for the best
i'm sure they've thought of everything
 
put it on the interrogate erik agenda
 
which part?
 
1:11 PM
@PaulWhite It's no doubt a thankless job
Thanks by the way!
 
@George.Palacios for?
 
@PaulWhite being a caretakers/curator of course
In all seriousness though, when was the last time you were actually thanked?
 
oh I see, well I signed up for that
 
my sympathies are with the users that care about site quality, but don't have the tools or support they need
@George.Palacios Joe thanks me every day. It's become his main role
 
1:15 PM
Well everyone needs something to do I suppose :P
 
idle hands and all that
 
well when he's not pulling cords
well
none of my tests from last night are repeatable this morning
i wonder what happened
 
Sadly, I'm not dead. You're all stuck with me a little while longer.
 
@ErikDarling sobered up?
 
@ErikDarling did you turn it off and then back on?
 
1:21 PM
@SeanGallardy deprecate heaps while you're still alive
@PaulWhite i do not drink during the week :(
@Taryn doing that now
 
@ErikDarling WHAT
 
it would be tremendously counterproductive
i don't have the same "one glass" willpower that you do
 
@ErikDarling It'd be nice, for sure. I don't think I have that super power yet...
 
I read a thing once that said a glass of wine a day was good for you
 
@ErikDarling Depends if you count it as "one glass" or "8 ounces". Glasses come in varying sizes and shapes....
 
1:24 PM
Rumours of counter-arguments persist, but I pay no attention to rumours
 
@PaulWhite I remember that! I think it was in the same rations kit as the cigarette a day one.
 
ignores comment based on rumour
 
@PaulWhite red wine, dark chocolate, and sunlight
@SeanGallardy one cigarette a day, especially in the morning, is good for... digestion.
 
Something like that..
 
There is no doubt we've stopped making changes for core users and your observations about how that's gotten worse ring true to me. And we've certainly seen negative feedback on meta (downvotes being the most trivial). The result may not be what you hope for, however. Often (and more often recently) I've heard colleagues dismiss meta feedback. Nobody wants to listen to relentless negativity. Hence this meta post written in the dead of night so I can finally sleep. As an intermediary between the community and the company, I'd like to help meta feedback be less easily ignorable. — Jon Ericson ♦ Aug 19 '18 at 15:47
If only SE had a concept of paying staff who work on the site. Getting paid might make doing unpleasant things more tolerable
 
1:28 PM
One might also reflect on the root causes of "relentless negativity"
 
@JoeObbish well then they'd have to bring in revenue via monetizing traffic or paid versions of their platform
 
Or ignore it, and coast along on the success of the efforts of others in the past
 
and we know they'd never do that
 
Until someone comes along and invents a better SO
 
as someone who works in software, my position is if you hate your end users then you need to find somewhere else to work
 
1:30 PM
yep no one ever said it would be easy
 
I just take that attitude personally
 
perhaps criticism doesn't form part of a safe space for SO staff
 
the way that your unhappy end users communicate doesn't matter. they aren't the professionals. you are. If an end user alerts me to a problem via smoke signals then that's just as valid as anything else
 
@PaulWhite Nah, it should be summarily dismissed because the crystal palace knows best.
Wait did I just project ;)
 
heh
 
1:33 PM
@JoeObbish I agree with this.
Just because we expect a certain level of decorum from respected members of the community doesn't mean you can just ignore the people who don't hold that same standard
Seems counter-productive
 
it's frustrating to feel powerless
doing development within hours of a tweet by a non-user certainly doesn't help
 
Trying to read the conversation, I have the nagging sense that I miss hidden messages.
 
it's all in the transcript
 
@ypercubeᵀᴹ You didn't get the magic decoder ring in your cereal this morning?
 
1:36 PM
Feeling like Sheldon; is this sarcasm? Is this?...
 
is that new?
 
@PaulWhite cheers for the details! I'll add something about plan invalidation. I considered using DBCC PAGE in my demo, but couldn't get it to work in a way that could easily be added to my answer - it results in just so much text. I still have the code I wrote for it, so perhaps I'll revisit it and see if I can encapsulate something about that in my answer. As you guessed, fn_dblog was easier and George asked about it directly.
 
@ErikDarling new without a trace flag yes
 
-__________________-
 
1:37 PM
Slot 0 Column 1 Offset 0x4 Length 4 Length (physical) 4
c1 = 1

Slot 0 Column 67108865 Offset 0x8 Length 0 Length (physical) 20
DROPPED = NULL
 
Thnx Paul
 
@MaxVernon Reminds me to mark your answer as accepted actually
 
@ErikDarling I'm distinguishing between things that have been maturing for a while, and those those arrive suddenly like OPTIMIZE_FOR_SEQUENTIAL_KEY
 
i understand
 
so what's up with the hashPages thing?
 
1:39 PM
^
 
terminology
data and index pages in buffer pool are known as hashed pages
 
is there hashing involved?
 
@JoeObbish Yes
 
why tho
 
bug fixes and performance improvements
 
1:43 PM
to have lists that are roughly equal in size and allow for fast lookups
 
I see
is there also a message for allocating 0 bytes for the columnstore object pool? :^)
 
@PaulWhite I'm probably looking at the wrong pages, but I didn't see that.
I used:
SELECT dpa.allocated_page_file_id
    , dpa.allocated_page_page_id
    --, dpa.*
FROM sys.schemas s
    INNER JOIN sys.objects o ON o.schema_id = s.schema_id
CROSS APPLY sys.dm_db_database_page_allocations(DB_ID(), o.object_id, NULL, NULL, 'DETAILED') dpa
WHERE o.name = N'DropColumnTest'
    AND s.name = N'dbo';
as my source for pages
 
@JoeObbish Only 4u
 
@JoeObbish there are some details about BUF structures etc. in blogs.msdn.microsoft.com/psssql/2010/03/24/…
there's probably a more focused resource out there that's not in a book
 
Some SQL Server versions (Enterprise for example) will do additional ramp-up reads. When SQL Server is first started each page read is expanded to 8 pages so the buffer pool is populated quickly with additional pages near the pages that are currently being asked for. This can reduce the time required to return the buffer pool to a warm/hot state. Once the commit target it reached the ramp-up behavior is disabled.
Is that the same as read-ahead?
 
1:52 PM
@MaxVernon I get:
Slot 0 Column 1 Offset 0x4 Length 4 Length (physical) 4
rid = 1

Slot 0 Column 67108865 Offset 0xf Length 0 Length (physical) 8000
DROPPED = NULL
Make sure you're looking at a data page
AND dpa.page_type_desc = N'DATA_PAGE'
May 28 at 6:50, by Paul White
@JoshDarnell My favourite reference for ramp-up reads is https://blogs.msdn.microsoft.com/ialonso/2011/12/09/the-read-ahead-that-doesnt-count-as-read-ahead/
does no one read the transcript
 
@George.Palacios no, not at all
 
Not from before I joined the room in general
 
read-ahead can happen at any time
 
@George.Palacios Well that's where you're going wrong
 
@PaulWhite nope
 
1:57 PM
Brb in 5 years - reading the entire transcript
3
 
@Taryn well you'll never get far with that attitude
 
@George.Palacios every time you laugh at something someone said, you owe them $5
 
@George.Palacios take a packed lunch and a warm rug
 
Oh boy
 
a flashlight and box of matches may also be useful
 
1:58 PM
a warm... rug?
 
@PaulWhite reading the transcript can be depressing
 
For the darkest days of the chat after 2010?
 
I18N error detected
 
@JoeObbish yes for the cold nights
 
did someone just insult the precious transcript?
 
1:59 PM
@Taryn not in here, you must be thinking of meta
or a different chat room
this room is relentlessly upbeat
 
R E L E N T L E S S
 
REEEEEEEEEEEEEEEELENTLESS
2
 
@PaulWhite rugs are for stepping on
 
@PaulWhite I just smacked myself. I clearly don't know enough about looking at DBCC PAGE output. Thanks again.
 

« first day (513 days earlier)      last day (176 days later) »