When Does a Carpenter Search Start – and When Does It Stop?

I don't have answers, but I have a lot of questions.


The Supreme Court's recent decision in Carpenter v. United States raises lots of fascinating and novel Fourth Amendment questions. In this post I want to focus on one important set of questions: When does a Carpenter search start, and when does it stop? I don't have a lot of answers to these question. But the questions are important to consider. This post, cross-posted at Lawfare, will explore the questions and will end with a set of hypotheticals to bring out the tough issues.

This is a long post, so here is a roadmap. I'll begin by explaining why the timing of Fourth Amendment searches was pretty easy before Carpenter, and I'll turn next to why Carpenter complicates the issue. I'll then discuss the stages of surveillance and explain why the timing of Carpenter searches is pretty important. After briefly covering the timing of searches suggested by the Jones concurrences, the next section goes through the language of Carpenter and argues that the opinion leaves the timing questions pretty open. I'll end with four hypotheticals designed to bring out different ways of thinking about when Carpenter searches start and stop.

I. Before Carpenter, the "When" of Search Law Was Simple

In traditional Fourth Amendment law, identifying when a search starts and stops is easy. That's true because Fourth Amendment law has always before been connected to a place or thing. There was always some specific place, thing, or person that was searched, and the search occurred when the information about that place or person was revealed. The place searched could be a house (as in Kyllo), a structure (the phone booth in Katz), a car (as in the Jones majority), or a person (like a Terry frisk). But it was always a specific and identifiable place, thing, or person – what the text of the Fourth Amendment refers to as the "persons, houses, papers, and effects" that the Amendment protects.

Identifying when a search occurs under this traditional approach is very simple. You just look at the moment when information from the place, thing, or person is revealed to the government. Figuring out the timing of that search is traditionally straightforward. For example, imagine the police break open a door to an apartment and rifle through its contents for an hour before leaving. Identifying when the search occurred is obvious. The search begins when the door is opened, lasts for an hour, and ends when the police leave. It's pretty intuitive.

II. How Carpenter Makes the "When" Question Difficult

The timing of a Carpenter search does not seem intuitive, however. Carpenter is unique among majority opinions of the Supreme Court in that it recognizes a right against the government collecting a set of facts in the abstract. And that unique kind of right has big implications for the timing of Carpenter searches.

Let me explain what I mean. The starting premise of the Carpenter opinion is that, at some point in the past, you wouldn't have expected the government to be able to collect lots of location information about a suspect. "Prior to the digital age," Carpenter states, "law enforcement might have pursued a suspect for a brief stretch, but doing so for any extended period of time was difficult and costly and therefore rarely undertaken." Quoting Justice Alito's Jones concurrence, Carpenter concludes that this means that there was in the past a reasonable expectation "that law enforcement agents and others would not—and indeed, in the main, simply could not—secretly monitor and catalogue every single movement" of a person over time."

Accessing CSLI is a search, Carpenter reasons, because that has changed. Because technology has now made that kind of surveillance easy and potentially very common, the law needs to step in and make that surveillance difficult and rare again. As I have written before, this is the theory of equilibrium-adjustment. When technology dramatically expands the government's power under an old legal rule, the thinking goes, the Court changes the legal rule to restore the prior level of government power. To ensure that the government doesn't have unlimited power to catalog your movements over time, Carpenter holds that a search occurred and a warrant was needed. New technology means that you no longer have a reasonable expectation of privacy in your movements over time; the law gives back to you the reasonable expectation of privacy that you once had. The law giveth what technology taketh away.

But when exactly does the search occur? The Court's phrasing on this point is fascinating. The majority writes, in a key section, that "[t]he location information obtained from Carpenter's wireless carriers was the product of a search." The product of a search. A search occurred at some point, such that at the end we can say that a search caused the information to be obtained. But exactly when?

I think this is hard to figure out because Carpenter holds, for the first time, that a search occurred without it being a taking of information from any particular place, thing, or person. A search occurred because it needed to have occurred to regulate a practice that needed to be regulated to keep the government from having too much power. But the gathering of information was not taken from a particular place or thing or person at a particular moment. Rather, the government simply ended up with too much information about someone. How it ended up with too much information isn't particularly relevant in the Court's view. The point is the result, not the process. Somewhere in the steps that ended up with the government getting all that information, a search occurred.

III. The Stages of Surveillance

Why does this matter? It matters because modern systems of records collection and access often have several different stages. It's often important to figure out which stage or stages is the one that counts as a Fourth Amendment search.

Let's break down how it might work when the government ends up with cell site location records. I will break down the process into eight stages, some of which will blend together in practice but that may be conceptually helpful to keep distinct. To be clear, the eight stages are highly stylized and somewhat arbitrary; this or different methods of surveillance can be described differently. But at least as an exercise, consider the following eight steps.

(1) Creation of the Record. The relevant record first must be created, either because it is necessarily created to deliver the product or services or because a computer was programmed to create the record for other reasons. For example, if a cell phone call is made from an account at 2:34pm, and cell site 2342837413 is used to connect the call, the network will have generated that information.

(2) Preservation of the Record. Next the record must be stored. Here the cellular provider decides that this particular record will be saved and preserved, either for its own business purposes or pursuant to a legal obligation.

(3) Creation of a Database. The cellular provider takes all of the similar records and puts them into a database with similar records. For example, if a person made five phone calls on a particular day, the provider might have a database file for that day that looks like this:

March 2 at 2:34pm connection to site 2342837413.

March 2 at 3:09pm connection to site 2342837413.

March 2 at 4:12pm connection to site 2342837413

March 2 at 4:18pm connection to site 1893472393.

March 2 at 4:51pm connection to site 2342837413.

(4) The Database File Is Sent to The Analyzing Agency. Acting pursuant to a court order, he cellular provider takes the database file and sends it to the government. The file is received by the government but not yet opened.

(5) The Database File Is Opened by The Analyzing Agency. The government agency opens the file and a person there examines the data.

(6) Another Database Is Obtained to Make The First Database File Useful. In this case, the initial file was a list of times and raw numbers of cell sites. The raw numbers don't tell you much yet, though, as you don't yet know where the cell sites are. (All you know is that the phone was using one cell site, then used a second, then went back to the first.) To get locations from the database you need a second file, one that associates cell site numbers with their physical locations. In our case, the cellular provider might then send on a second file that looks like this:

Site 2342837413 is located at 6th Street and Elm Road.

Site 2362383929 is located at 23rd Street and Elm Road.

Site 1283839292 is located at 96th Street and Johnson Place.

Site 1893472393 is located at 47th Street and Baker Avenue.

(7) The Databases Are Combined. At this stage the two databases are put together to make the information useful in a way that can be easily queried. In this case, putting the entries together might look like this:

March 2 at 2:34pm connection to the site at 6th Street and Elm Road..

March 2 at 3:09pm connection to the site at 6th Street and Elm Road.

March 2 at 4:12pm connection to the site at 6th Street and Elm Road.

March 2 at 4:18pm connection to the site at 47th Street and Baker Avenue.

March 2 at 4:51pm connection to the site at 6th Street and Elm Road.

(8) The Combined Database Is Queried. With the database assembled, it can now be queried. For example, a government analyst may query the database for the times that the phone was in the area of Baker Avenue, with the response being March 2 at 4:18pm. Or the query might be for the location of the phone after 4pm on March 2, which would return the three points at 4:12pm, 4:18pm, and 4:51 pm. Or the query might be for any times when the phone was in the area of 96th Street and Johnson Place, which would return a null response.

Those are the eight stages. The big question: When does the search start, and when does it end?

IV. How the Jones Concurrences Answered the "When" Question

This problem first arose with the concurring opinions in United States v. Jones. In Jones, the government placed a GPS device on a car the suspect was driving and monitored the car's location for 28 days. The Jones concurrences introduced the basic premise of Carpenter that a search occurs because the monitoring simply gathered too much information. But the concurrences had different ways to describe when the search occurred.

As I noted in The Mosaic Theory of the Fourth Amendment, the two concurring opinions in Jones both looked beyond the initial data acquisition stage, albeit in somewhat different ways, for when a search occurred:

Justice Alito's opinion in Jones looked to whether a person reasonably expects others to "secretly monitor and catalog" a person's movements. Justice Sotomayor asked "whether people reasonably expect that their movements will be recorded and aggregated" in a manner that creates the mosaic. Cataloging and aggregating are verbs that describe subsequent analysis instead of initial collection. These phrases suggest that the mosaic theory requires some step beyond the acquisition stage.

If so, courts will need to determine what kinds of post-acquisition conduct are required to create a mosaic. Imagine the government collects a great deal of information but never combines it into a single database. Has a mosaic been created? Or imagine the evidence is collected into a database but never analyzed. Does that cross the line? If some analysis of the evidence is required to trigger the mosaic, what kind of analysis counts? Does any analysis suffice, or is there some threshold of sophistication or computational complexity before the mosaic line has been crossed?

Carpenter seems to have adopted the basic mosaic approach of the Jones concurrences. Given that the Jones concurrences suggest that some later-stage analysis is at least part of the search – some cataloging or aggregation is required—it raises the possibility that Carpenter searches require a similar range of steps.

V. What Carpenter Says About Timing – And What It Doesn't Say

That brings us, finally, to Carpenter itself. The Carpenter majority opinion gives us a few clues on the timing of when a search at least starts. But the clues don't seem particularly precise or consistent. And while they talk about when the search may start, they don't seem to tell us when the search ends.

That's not necessarily criticism of the majority, to be clear. The timing question wasn't presented to the Justices because it didn't matter in that case. But if we're looking ahead to how to apply Carpenter, the timing clues left in the opinion start to matter.

So let's take a look. Some language in the majority opinion say that a search occurs when the government "accesses" a database that contains the relevant records. To me, accessing seems to mean looking through the database to get useful information. Here are a few examples:

(1) "This case presents the question whether the Government conducts a search under the Fourth Amendment when it accesses historical cell phone records that provide a comprehensive chronicle of the user's past movements."

(2) "Allowing government access to cell-site records contravenes that expectation."

(3) "With just the click of a button, the Government can access each carrier's deep repository of historical location information at practically no expense."

(4) "Accordingly, when the Government accessed CSLI from the wireless carriers, it invaded Carpenter's reasonable expectation of privacy in the whole of his physical movements."

(5) "We decline to grant the state unrestricted access to a wireless carrier's database of physical location information."

Other parts of the opinion refer to the relevant stage as "acquiring" the records instead of "accessing" them. The word "acquiring" can have different meanings. On the whole, though, that word seems to suggest a relatively early stage of merely coming into possession of the records even if they are not actually examined or queried.

Here are some of the passages that refer to acquiring or similar language instead of accessing:

(1) "The case before us involves the Government's acquisition of wireless carrier cell-site records revealing the location of Carpenter's cell phone whenever it made or received calls."

(2) "The Government's acquisition of the cell-site records was a search within the meaning of the Fourth Amendment."

(3) "Having found that the acquisition of Carpenter's CSLI was a search, we also conclude that the Government must generally obtain a warrant supported by probable cause before acquiring such records."

(4) "The Government's acquisition of the cell-site records here was a search under that Amendment."

(5) "Before compelling a wireless carrier to turn over a subscriber's CSLI, the Government's obligation is a familiar one—get a warrant."

So which is the search, acquiring the records or accessing them? Possessing them or making use of them? I'm not entirely sure.

Beyond the possible acquisition/access distinction, there are two important additional uncertainties on how Carpenter deals with the timing.

First, Carpenter assumes that all of the records have been combined into a database that can be queried to get all the relevant information. It doesn't say how the law should apply when the relevant information has to be obtained from combining databases. Recall stages 6 and 7 above in Part III. The first database didn't actually have the locations of the cell towers. It had only raw numbers. The second database was needed to link site numbers to actual places.

I wonder, in this hypothetical, would just the first database count as "location information" for Carpenter purposes? Is either acquiring or accessing that database alone a search? There is no actual "location" information in that database. It is just meaningless numbers. But it's a good step towards having location information. Is that enough? Does it matter how easy or hard it is to combine those meaningless numbers with the location information? For example, does it matter if the second database file, with the links between site numbers and location, is a closely-guarded secret versus if it is something that can be looked up on the public web or is given out relatively freely by the provider? You can see why it matters. Going back to the eight stages of surveillance in Part III, the answers tell you when a search starts: Does it begin at stage 4, 5, 6, 7, or 8? It's hard to say, I think.

The next uncertainty is when the search ends. Let's say you identify a stage when a search starts. Does the search end there, too? Or is the search ongoing for some set of the stages of surveillance? This isn't a hard problem in traditional Fourth Amendment law because you would traditionally say that once data has been searched, further examination is not a search again. As the Supreme Court stated in Illinois v. Andreas, 463 U.S. 765, 771–72 (1983), "once the police are lawfully in a position to observe an item firsthand, its owner's privacy interest in that item is lost." Observation can be a search, but searching once eliminates the expectation of privacy. Or at least that's the traditional view of the Supreme Court's caselaw.

But does Carpenter change that? If Carpenter searches are about accessing information, you might get new searches at different stages of the surveillance as more information is obtained. You might have one search at one stage and a second or third search later. If the government has a warrant at the outset, of course, the warrant should suffice for all the steps. A warrant authorizes as many distinct searches as are needed to search the described place for the described evidence. But what happens if an exception to the warrant requirement applies at one step but not another? Say you can get to a particular stage of surveillance where a search has occurred but a warrant was not necessary because an exception to the warrant requirement applied such as exigent circumstances. Next say that the exception no longer applies going forward. Are subsequent stages of surveillance new searches that now require a warrant to execute? Or is the search over from the earlier stage of surveillance so the subsequent stage can occur without Fourth Amendment oversight?

VI. Four Closing Hypotheticals

Here are a four hypotheticals to bring out these uncertainties and make the issues concrete. To be clear, these aren't exam questions where I know the best answers and I'm wondering who among you readers can find them. Rather, these are uncertain issues under Carpenter that I'm not sure how courts will answer. I'm curious to know what you think of them.

(1) "Has David Ever Been To California?" Say David has been charged with killing a man in New York. The exact time of the death is unclear, but the crime is thought to have happened in recent weeks. David's alibi is that he was in California all of the last month and that he could not have committed the crime in New York. Investigators want a representative from David's cell provider to testify at trial to answer just one question: Whether, based on a review of the last month of David's CSLI records, David's phone connected to any cell sites in California at any time in the last month. Is obtaining that testimony a Fourth Amendment search that first requires a warrant, and if so what kind of probable cause showing is required for it?

(2) "Just Tell Me If When There Was a Match." Imagine the government is investigating a robbery conspiracy. Investigators want the provider to disclose the times and dates in the last year when the cell phones used by the four suspects were connected to the same cell tower within 15 minutes of each other. Investigators don't want to know where the cell sites were located. They only want a list of times, if any, when all four cell phones were using the same cell site (without knowing where that was). An employee at the provider takes its databases and combine them to look for matches. He responds to the government with a list of times, and those times happen to match to the dates of known robberies. Was it a Fourth Amendment search to obtain that list of times? If so, do all of them have standing to challenge the search?

(3) "The Emergency Passed." The police are investigating an ongoing kidnapping. In an effort to find out where the kidnapper is, the police ask the suspect's cellular provider to disclose the whereabouts of the suspect's phone for the last month. The provider does so without a warrant in light of the exigent circumstances of needing to find the kidnapper right away. The government gets the location information covering one month under the exigent circumstances exception to the warrant requirement. The kidnapper is not found, however. A year later, investigators in a different case come to think that the same suspect may have also committed a string of burglaries around the same time. They want to examine the location information that was originally obtained to find the kidnapper, but this time they want to analyze the records to see if the locations matched the burglaries. Is this later examination another search, and if so does it require a warrant?

(4) "We'll Query the Database in An Emergency" Congress enacts a terrorist surveillance program that requires every cellular provider to hand over all CSLI to the government once a week. The government is required to save all the CSLI files for all time without looking at any of the data. When an emergency amounting to exigent circumstances occurs, however, the government can query the CSLI database to obtain location records relating to particular suspects. Has a search occurred if no emergency has occurred yet and the records have been obtained but not opened or analyzed?

NEXT: Marquette University "Breached Its Contract with Dr. McAdams When It Suspended Him for Engaging in Activity Protected by the Contract's Guarantee of Academic Freedom"

Editor's Note: We invite comments and request that they be civil and on-topic. We do not moderate or assume any responsibility for comments, which are owned by the readers who post them. Comments do not represent the views of Reason.com or Reason Foundation. We reserve the right to delete any comment for any reason at any time. Report abuses.

  1. Never – there ain’t no search like a Carpenter search, ’cause a Carpenter search don’t stop.

  2. I strikes me that legal convolutions have become ever-greater with the advancement of technology, and that isn’t likely to stop. I suggest a need to disentangle search questions from the specifics of technology, and the happenstance of everything else. Permit me to suggest the following 2 points:

    1. It’s counter-intuitive to suppose that the government can, without initiating a search, put a person under surveillance in the hope of discovering probable cause. Isn’t that the moment when the target’s personal security actually diminishes? If the government does that, why isn’t that the initiation of a search without a warrant? Presumably the answer will come back, “No, it’s not a search, because tall stacks of precedents say it isn’t.” That may dodge the question, but doesn’t answer it.

    2. If, instead, the government had to wait for probable cause to come by happenstance, or in the course of other law enforcement activity, then the answer to Orinn’s question would seem easier. A search begins at the moment law enforcement officers turn their attention toward a particular person with the aim of proving that person has broken a law. Let technology be what it may, making that moment the marker for when a search begins might prove a durable standard. And if probable cause is already in hand, how hard will it be to get a warrant? And if probable cause isn’t in hand, what the hell is going on when law enforcement singles someone out as a target?

    1. “It strikes me . . .”

    2. The fourth amendment only protects against unreasonable searches. Following someone around (in public) may well be a search, that doesn’t say anything about whether it is reasonable or not. And the case law has come down to saying that such following is not unreasonable (mostly because people have no reasonable expectation of privacy in their public movement).

      1. The problem is that if the scope of searches that are “reasonable” without a warrant becomes too broad, it will entirely swallow the warrant requirement.

        In my opinion, warrantless searches should be considered per-say unreasonable unless certain conditions are met that are both very narrow, and very clear. Current case law does not meet this criteria.

      2. With regard to stuff seen in public, I had supposed that a certain amount of that might be encompassed within my reference to happenstance. If, in public, something which suggests probable cause comes to the attention of a law enforcement officer, it’s reasonable that further scrutiny would follow. But it wouldn’t be happenstance?or reasonable?if the law enforcement officer was already out looking for that particular target without already having probable cause.

        I’m a bit nonplussed by the role “expectation of privacy,” has been assigned. Making reasonable expectation of privacy the threshold for protection against unreasonable searches opens the field too much. On what basis, other than precedent, should society assume that everything not within a narrow definition of expectation of privacy is fair game? Why presume the Constitution means that all of public life is reasonably thrown open to every intrusive exaction that law enforcement can muster, without any judicial oversight? Isn’t that kind of a totalitarian vision?

        1. Another point about expectation of privacy. Putting that together for consideration jointly with what constitutes a reasonable search makes the whole enterprise unhappily recursive. The answer to what privacy might reasonably be expected will be notably affected by what searches might be authorized. You can hardly make the former a standard for gauging the latter, unless you are prepared sufficiently to define the scope of reasonable searches to also make the privacy expectation definable. And if you do that, you really don’t anymore have a need to judge what privacy is reasonable, because that question has been answered by the boundary affixed to limit searches.

  3. Regarding your 4th closing hypothetical:

    I would have issues with the whole setup, but at the top of the list is this, the government should only be allowed to use the exigent circumstance exception to access records that were generated after the exigent circumstance came in to being and any access at all ceases to be justified once the exigent circumstance ceases to exist.

  4. Pedantry: Steps 4 and 5 are the same step. “Query” is a term of art; combining databases is a form of query, and it would generally be very stupid not to select the ones you want to look at while doing that.

    It seems to me that the search begins when the government demands the records, and ends when the records are produced, but that’s just firing from the hip.

  5. The characterization of the hypothetical cell site ID numbers in the hypothetical stage three database as “meaningless numbers,” and thus categorically not location information when asking whether accessing or acquiring that database would constitute a search, strikes me as incorrect. Those cell site ID numbers are unique, persistent labels which refer to a specific, unchanging physical location in the exact same way that a traditional address (street number and name) or the names of two crossing streets do. The cell site ID numbers are meaningless only to the extent that a particular person viewing them doesn’t have the information necessary to understand them. The same can be said, for example, of latitude and longitude coordinates. The vast majority of people, given a set of such coordinates, wouldn’t have even the foggiest idea where they referred to in the real world, yet such coordinates are, literally by definition, location information.

    Even entertaining the notion that such labels are not location information, and thus accessible without a warrant, seems like an invitation to create a potential loophole in Carpenter.

  6. I believe the Answers to hypothetical questions 1 and 2 is easy to answer.
    The “Government Agent” doctrine would apply, so a search warrant would be necessary.
    A person, acting on behalf of the government, that conducts a search, is no different than if the police conducted the search themselves. This is long established 4th Amendment Law. In both case in 1 and 2, the Employee is in fact conducting the search now barred by Carpenter.

    As for Hypothetical #3, Although a bit trickier, I think a warrant there is also required. In the normal course of action with physical information, once a search has proven fruitless, the physical items are returned. Thus, a second search of items (physical) that were seized under the Exigent Circumstance” exception would not or could not really happen (in most cases) Since the carpenter Decision basically put electronic information held by a third party on the same footing as Physical papers, logic would dictate that a warrant would be required for the second search. However, I do not believe there is relevant case law to support either position.

    As for Hypothetical #4, I believe the relevant part there is the act of Congress. If the acquisition of the records is deemed constitutional, then by extension, they are the possession of the government, and therefore, a warrant would never be needed… But I believe the law itself would not pass Constitutional muster.

    Just my opinions here – Do not take any of it as legal advice!

  7. Hypo (1) Combining the analysis in United States v. Ceccolini, 435 U.S. 268 (1978) with section 1.8(b) of LaFave’s “Search & Seizure” (5th ed. 2012), no ? the government cannot do this without a warrant because it is a private search at the government’s specific instigation.

    Hypo (2) Just a variant of Hypo (1), the government is still using a private party to circumvent the Fourth Amendment and do something it cannot lawfully do without a warrant. Also, Lee Tien, Privacy, Tech and Data Mining, 30 Ohio N.U. L. Rev. 389, 408-09 (2004) (arguing that data mining is a search because “underlying patterns or associations are not the same as the surface facts in the database, and that these patterns may themselves be private.”).

    Hypo (3) Assuming that the investigators who were working on the kidnapping still have possession of the records that may now be relevant to the later burglary investigation, Lafave at section 5.3(b) of his “Search and Seizure” (5th ed. 2012) (particularly the cases at n.76 and n.86) lays out a framework for the analysis of having a second look at evidence obtained under an exception to the warrant requirement. Yes, the government can take a second look without a warrant.

  8. Hypo (4) No, the government cannot seize now as a matter of routine and search later. LaFave at section 2.7(e), pp. 992-93; and Paul Ohm, The Olmsteadian Seizure Clause: The Fourth Amendment and the Seizure of Intangible Property, 2008 Stan. Tech. L. Rev. 2, 74 (2008) (arguing that while such a practice may be lawful under the national security exception to the 4A, the initial seizure is nonetheless a 4A violation in criminal investigations that cannot be cured by post-hoc authorizations).

  9. It may be that I am a bear of very little brain, and the answer is above my head, but I would say that the search begins at the date of the oldest record to which the investigator obtains access. I don’t care what he’s looking FOR. I care mightily about what he is able to look AT. Giving attention to the soft-shoe dance of data manipulation is misleading. It’s a smoke screen for the investigator to hide behind. He has grabbed a pile of stuff, and he can see the oldest item in the pile, unless he has gone to great lengths to render it invisible.

    Not a lawyer. Software guy since 1962. Seen a lot of stuff I shouldn’t have, including Top Secret.

    1. I’m not a lawyer, either. But isn’t caring what the investigator is looking for a notable part of determining whether he is looking reasonably? If you ask the investigator, “What are you looking for?,” and the answer comes back, “A little of this, a little of that, whatever might show the guy committed a crime,” isn’t that by 4A definition unreasonable?

    2. Analogy: The time when the search takes place is equivalent to a query of remote data.

      The “scope” of a warrant is equivalent to a SELECT statement. It narrows a query but doesn’t self-execute. This is where you’d choose the dates of the records.

      But as I understand it, the question is when the query takes place. Does it begin when you start the local application, when you launch the SQL service, when you transmit the query, when it’s received, when it’s authenticated, when execution begins, or some other time? When does it finish?

      That’s not entirely philosophical, because there are laws here and there limiting when warrants may be executed (i.e. forbidding night-time raids without good cause shown). I suppose there might be staleness arguments to be made some day, and probably other stuff I’m not thinking of.

Please to post comments